# askr docs Every AI model, one balance, fully documented. Start with a section, or search. Verified against the code on 2026-09-15. ## Three ways in, one balance - The workspace: Chat, code, images and video in the browser. Sign in with your email and start. No key needed. - The CLI: Three commands in a terminal and you are talking to any model, with your key remembered. - The API: An OpenAI-compatible endpoint. Point Cursor, Continue, Aider or the OpenAI SDK at it with a base URL and a key. ## What askr adds to the models - askr Mind: What you told one model is there for the next: a short memory per chat and per room, and one about you, first in every prompt and yours to edit, pin or forget. - askr Collab: A room where people work the same models together out of one purse, with threads, Compare and a public link. A chat of your own is a room of one. - askr Skills: Ten chats behind Ask, each built for a job: sources, drafts, worked steps, a fix card, results on bars, calculators, ideas, a compare table. ## Start in four steps 1. Sign in with your email. There is no password. You get a six-digit code by email and enter it. How sign-in works. 2. Add credits. Send crypto to an address issued for that one deposit. One dollar is 1,000 credits. Adding credits. 3. Make a request. In the workspace, the CLI or from your own code. Your first request. 4. Watch what it cost. Every request shows its cost in credits when it finishes, and the wallet lists all of them. What a run costs. ## Where things are - Workspace: https://heyaskr.ai/chat - Wallet, keys and activity: https://heyaskr.ai/wallet - Model catalog and prices: https://heyaskr.ai/models - API base URL: https://heyaskr.ai/v1 - Help: ask@heyaskr.ai ## What a stack of subscriptions costs For comparison: the monthly bills askr replaces. One balance, drawn down only when you run something, against these renewing whether you use them or not. > These pages describe what the platform does today. Each one carries the date it was last checked against the code, and anything the platform knows (model ids, prices, deposit assets) is read live rather than typed here. Source: https://heyaskr.ai/docs --- # Create an account and sign in How do I get an account, and what does signing in look like? Verified against the code on 2026-09-15. An account is an email address. Signing up and signing in are the same act: enter your email, we send a six-digit code, you enter the code. There is no password to choose or remember. 1. Go to the sign-in page. Open https://heyaskr.ai/signin, or press Start asking anywhere on the site. 2. Enter your email. We send a code to it. If the address has never been used, the account is created the moment the code is verified. 3. Enter the code. Six digits. It expires ten minutes after it was sent, and a code only works for the address it was sent to. ## Limits that protect the account - Five codes per email address per hour. Ask for a sixth and you wait. - Five wrong guesses per code, then that code is dead and you request another. - A session lasts 30 days. Sign out everywhere in the wallet ends every session at once. ## What we store Codes and session tokens are stored only as SHA-256 hashes. A copy of the database would not contain a working code or a session anyone could replay. Security has the detail. ## If the code does not arrive - Check spam. The sender is no-reply@heyaskr.ai. - Wait a minute before requesting another; the newest code is the one that counts. - Still nothing after a few tries: email ask@heyaskr.ai from the address you are trying to use. Source: https://heyaskr.ai/docs/sign-in --- # Add credits How do I fund my balance, and when does the credit land? Verified against the code on 2026-09-24. You add credit by sending cryptocurrency to an address issued for that one deposit, or by card through Stripe (Pay by card, below). One dollar is 1,000 credits either way. The list of assets below is read from the crypto payment processor as it is right now. 1. Open the wallet. https://heyaskr.ai/wallet, then Add credits. 2. Choose an asset and an amount. Pick from the list below and enter the dollar amount. The quote shows exactly how much of the asset to send. 3. Send the exact amount to the address shown. The address is for this deposit only and is good for the countdown shown next to it. Send the quoted amount on the named network, on top of whatever your wallet charges to send it. 4. Wait for confirmation. Once the processor marks the payment complete, credits appear on your balance and a line appears in your activity. Times vary by network: seconds for fast chains, longer for Bitcoin on-chain. (The live deposit asset list is rendered here. JSON: /api/deposits/assets) ## Rules worth knowing before you send - Minimum $5 and maximum $10,000 per deposit. - Send the exact amount quoted, in one transaction. A short payment is credited pro rata for what arrived; an overpayment is credited for all of it, up to four times the quote. - The rate is fixed when the deposit is created, not when it confirms. - Crypto transfers are irreversible. Funds sent as the wrong asset, on the wrong network, or to an old address are usually unrecoverable. - Network fees are yours. They are not credited. - You are credited for the full amount you paid, at 1,000 credits per dollar. The processor takes about 1% and askr absorbs it today; askr adds no fee of its own. > The full detail, including what happens to a deposit that arrives after its window, is in Deposits. ## Pay by card Card payments go through Stripe, and each purchase is sold through Link, Stripe's checkout and customer brand, as the merchant of record: Link sells it on askr's behalf and takes the payment, and askr supplies the credits. In the workspace open Settings, then Balance & deposits (https://heyaskr.ai/chat/settings/balance), where Pay by card sits beside the crypto deposit, or pick the Card tab in Top up. Pick an amount, tick the box, press pay, and pay on Stripe's checkout page. Then you come back to askr and the credits land as soon as Stripe confirms the payment. Your card details go to Stripe and never reach askr. - Amount: From $5 to $500 a payment. Pick $5, $10, $25, $50 or $100, or type your own amount, to the cent. - Rate: 1,000 credits a dollar, the same as crypto: $25 is 25,000 credits. askr adds no fee and carries Stripe's. Prices include any tax: the price you pick is what you pay. - Before you pay: In askr's checkout step you tick a box to say you want the credits added straight away and understand you cannot cancel once they are on your balance; pay stays off until you do. Stripe's payment page then asks you to accept the terms. Your bank may ask you to confirm the payment. - Receipt and statement: Your receipt comes from Link, and the charge shows on your card statement as LINK.COM* HEYASKR.AI. askr also emails you a confirmation of the credits and what you agreed to. - When it lands: As soon as Stripe confirms the payment, usually within seconds. A checkout left unpaid lapses after 30 minutes and nothing is charged. - No cancelling once it lands: Card credit is supplied at once, and by ticking the box you ask for that. Once the credits are on your balance, askr does not refund the purchase. Link, as the seller, decides refunds of the payment under its own terms and applies a cooling-off period where the law requires one. - Card credits cannot be sent: They stay on the account that bought them. The balance shows how much of it came from card payments, and Sendable now in the send dialog leaves it out. A send that would need card credits is refused with W912. - Spend order: Card credits are bought credits, so the holder discount applies to them. A run spends airdropped and given credits first, then card credits, then credits bought with crypto, so the credits you can send last longest. - Spent credit: Not refundable. It paid for model work askr has already paid the provider for. - A failed run: Refunded in credits, never to the card. A run that fails is not charged, as What you are charged when sets out, and what it held comes back to your balance. - Disputes: A chargeback for credit you used breaks the terms, and askr contests it with the account's usage record. While a dispute is open the account cannot send credits or pay by card again (W913). If the dispute is lost, up to the credits that payment bought are taken back from the card credits on the balance, and whatever that leaves uncovered is owed: until it is paid, the account cannot run models or send credits (W302). A crypto top-up pays what is owed first; a card payment waits until nothing is owed. - Refunds: If Link, Stripe or askr refunds a card payment, in full or in part, up to the credits the refunded amount bought are taken back from the card credits on the balance, and whatever that leaves uncovered is owed, with the same pause as a lost dispute (W302). - Fraud: askr may refuse, freeze or reverse a card purchase it suspects is fraud. The rules in full are in the terms, and what Stripe and Link receive is in the privacy notice. Source: https://heyaskr.ai/docs/add-credits --- # Your first request How do I run something, in the workspace, the terminal, or from code? Verified against the code on 2026-09-15. Pick whichever fits. All three draw from the same balance and the same catalog. ## In the workspace Sign in, open https://heyaskr.ai/chat, choose a model in the picker and type. The price of the turn is shown before you send and the cost is shown when it finishes. The workspace. ## In the terminal The CLI needs Node 18 or newer and nothing else. It remembers your key and your model. ```bash curl -fsSL https://heyaskr.ai/install.sh | sh askr login askr ``` ```powershell irm https://heyaskr.ai/install.ps1 | iex askr login askr ``` askr login asks for a key from your wallet and saves it. askr on its own opens a conversation. Quote a question to ask once and exit: askr "what changed in HTTP/3?". The CLI. ## From your own code Make a key at https://heyaskr.ai/wallet#api. The base URL is https://heyaskr.ai/v1 and the shape is the OpenAI chat completions API. ```bash export ASKR_KEY="askr_live_..." curl https://heyaskr.ai/v1/chat/completions \ -H "Authorization: Bearer $ASKR_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-5.6-sol", "messages": [{"role": "user", "content": "Say hello to askr."}] }' ``` ```python from openai import OpenAI client = OpenAI(base_url="https://heyaskr.ai/v1", api_key="askr_live_...") resp = client.chat.completions.create( model="gpt-5.6-sol", messages=[{"role": "user", "content": "Say hello to askr."}], ) print(resp.choices[0].message.content) ``` ```javascript import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://heyaskr.ai/v1", apiKey: "askr_live_..." }); const resp = await client.chat.completions.create({ model: "gpt-5.6-sol", messages: [{ role: "user", content: "Say hello to askr." }], }); console.log(resp.choices[0].message.content); ``` The response is the OpenAI shape plus one field of ours, askr.credits_charged, so your code knows what the call cost without a second request. ```json { "id": "chatcmpl-6f1e...", "object": "chat.completion", "model": "gpt-5.6-sol", "choices": [ { "index": 0, "message": { "role": "assistant", "content": "Hello, askr." }, "finish_reason": "stop" } ], "usage": { "prompt_tokens": 12, "completion_tokens": 4, "total_tokens": 16 }, "askr": { "credits_charged": 0.0675 } } ``` ## Switch models by changing one string Every chat model in the catalog works with the same request. Ids are exact, so copy them from the catalog rather than guessing: claude-sonnet-5, gemini-3.7-flash, deepseek/deepseek-v4.1-flash. - Chat completions, in full: Every field we read, every limit, every error. - Streaming: Tokens as they are generated, and what a cancelled stream costs. - Tools: Cursor, Continue, Aider, the OpenAI SDKs, LangChain. Source: https://heyaskr.ai/docs/first-request --- # Frequently asked The short answers, with links to the long ones. Verified against the code on 2026-09-24. ## Do I need a subscription? No. You fund one balance and pay for the model usage you actually consume. Nothing recurs, nothing renews, nothing needs cancelling. ## How do I pay? With cryptocurrency, sent to an address issued for that one deposit, or by card through Stripe. One dollar is 1,000 credits either way. The assets accepted and the card rules are on Add credits. ## How much does it cost? Each request is metered at the model's own rate. A short chat on a cheap model costs a fraction of a credit; a video clip costs hundreds. Prices are on Models and the cost of every request is shown when it finishes. There is no platform fee in the first version. ## Do my credits expire? No. Credit does not expire. It buys model usage only, earns nothing, and cannot be withdrawn or converted back. Refunds and withdrawals. ## Which models can I use? Hundreds, across chat, code, images and video, from OpenAI, Anthropic, Google, DeepSeek, xAI, Qwen, Meta and more. The catalog is live and says which run on the API and which run in the workspace. ## Is my data private? Nothing you type trains a model. Messages and replies are not stored; a conversation exists in the browser tab you typed it in. Images and clips you make are kept in your library so you can get them back. Privacy. ## Can I send credits to a friend? Yes, to any @ on askr, from Settings, Balance & deposits in the workspace. Credits you bought with crypto or were given move; card credits, promo credits and the holder discount stay with you. The smallest send is 100 credits and there is no cap on how much you send; credits from a deposit wait 24 hours, and a send above 5,000 credits asks for a code by email. Sends are final; askr can reverse a mistaken or fraudulent send on request. They stay credits on both sides and cannot be redeemed for money. Send credits. ## Can I build on it? Yes. The API is OpenAI-compatible: change a base URL and a key and existing code works. The CLI is the fastest way to try it. ## Is there a token? Yes, ASKR. It is access, not a revenue claim: holding it lowers the askr fee and unlocks access bands. The token. ## Where do I get help? ask@heyaskr.ai, or the channels on Community. Source: https://heyaskr.ai/docs/faq --- # Chat How does the workspace work, and what are its limits? Verified against the code on 2026-09-24. The workspace is the browser side of askr: https://heyaskr.ai/chat. Sign in, pick a model, type. No key, no setup. Every model in the picker draws on the balance you funded. ## How a turn works 1. Pick a model. The picker shows what one typical turn costs on each model (1,000 tokens in, 500 out), cheapest first. Switching keeps the conversation, so you can hand the same question to a second model. 2. Type and send. The reply streams in. Stop it early and you pay for what was generated, nothing after. 3. See the cost. Each reply shows the credits it cost. The wallet lists every turn. For a question that needs today's facts, turn on Search the web with the globe beside the paperclip: the model searches, reads and cites its sources, at 5 credits a search. Ask opens a chooser: the plain chat, or one of ten skills built for a job, each a chat with its own screen (askr Skills). What you tell one model is there for the next, on the account, under your control (askr Mind); and a chat is a room of one, so it can hold threads, compare models side by side, and take people in (askr Collab). To give the model a file, use the paperclip, or drop the file on the chat: pictures, PDFs, text and code, Word, PowerPoint and Excel, audio and video. Uploading is free; a file costs tokens only on the messages that carry it. A model that cannot read a file itself gets its words, so a PDF or a document works on every model. Attach files has the limits, which models take which kinds, and what each costs. ## Limits - Messages kept: Up to the last 400 are sent to the model (20 on a model with a small window), oldest dropped first to fit - Message size: One message can fill the whole turn (48,000 characters on a model with a small window) - Conversation size: Up to the model's own context window, less room for the reply: about 1 million tokens on the largest models, such as Kimi K3, and never less than 120,000 characters. Older turns fall off, the chat continues - Reply length: Up to 32,768 tokens per turn, thinking included; less if the balance cannot cover it, or if the model itself takes less - Pace: 20 turns per minute - Deadline: 4 minutes per reply The API allows longer replies and conversations. When you outgrow the workspace, the CLI is the next step and takes a minute to set up. ## Speak instead of typing, and hear the reply A mic sits beside the paperclip in the composer. Tap it to listen and tap again to stop; sending stops it too. What you say is appended after what you typed, with the words still being guessed shown as they come and settled as you go, and a pause to think does not end the session. Four bars beside the mic move with your voice while it listens. Every finished reply has a read-aloud link that reads it in your browser's own voice: code is skipped with a word, links are read as their text, and one reading runs at a time. Both run inside your browser. Chrome, Edge and Safari carry the speech engine, so the buttons appear there and stay hidden in Firefox. No audio is sent to askr, nothing is recorded, and neither costs credits. The first tap asks for the microphone; if you have said no, the note tells you where the switch is on your platform. ## What stays and what does not Your chats are kept on your account, so they follow you across devices and are there when you come back. Each one is named by what it asks; the Chats tab of your account lists them and lets you download or delete any of them. Nobody at askr reads them and nothing you type trains a model; the privacy page says exactly what is stored. Images and video clips you make are kept in your library. ## If a model refuses to answer A reply with no content is treated as an error, not an answer, and is not charged. A model whose provider is not currently serving it says so before the turn runs. Listed but refused. Every error in the workspace ends with a code; the list says what each one means. Source: https://heyaskr.ai/docs/workspace/chat --- # Search the web How does the globe in the composer work, what does a search cost, and which models can do it? Verified against the code on 2026-09-23. A model on its own knows nothing after its training cut-off and cannot open a page. Turn on Search the web and, for that turn, it can: it searches, reads what it finds, and answers with numbered sources under the reply. Search and read is all it can do; it cannot act on anything it finds. ## Turning it on The globe sits beside the paperclip in the composer, in Ask and Code. Tap it and it stays on for that chat until you tap it again; each chat remembers its own setting. The estimate beside Send grows by the most the searches can add, and on a phone a line under the tool row says the same: Search the web is on, up to 15 cr search. On a model that cannot search, the globe stays in place but off, and hovering or focusing it says why: cannot search the web. Pick a model that can. Image and video models never offer it. ## What happens in a turn 1. The model decides. It is given two tools, web_search and fetch_page, and told today's date. It searches when the question needs current, specific or verifiable information and answers directly when it does not, so a turn with the globe on may run no search and cost nothing extra. 2. Searching the web for .... Each search returns the most relevant passages from up to 5 pages, with their titles and addresses, trimmed to about 3,000 tokens. The model can limit a search to the past day, week, month or year. The line above the streaming answer shows the query as it runs. 3. Reading .... A page you named, or one the model needs in full, is read by askr's server as plain text: the main content, trimmed to about 2,000 tokens. The line shows the site. A page whose robots.txt forbids automated reading, a private or internal address, and anything that is not HTML or plain text are refused. 4. The answer. Everything a tool returns reaches the model as data between markers, with a standing instruction that nothing inside them is an instruction, so a page that says "ignore previous instructions" is quoted, not obeyed. The model cites what it uses as [1], [2] and so on. - Searches: At most 3 a turn - Pages read: At most 2 a turn - Model calls: At most 4 a turn. The last is made without tools, so the model has to answer with what it has. - Pace: 6 searching turns a minute per account, workspace and API together - A search that fails: The model is told, and answers from what it knows, saying so if it matters. It still counts as a search. ## What it costs - 5 credits a search. Charged for every search that ran, whatever it found. - The text the model reads, as input tokens. Search results and pages go to the model like anything else in the conversation, at the model's own input rate, so a search on an expensive model costs more than the same search on a cheap one. Reading a page costs nothing beyond its tokens. Before you send, the estimate says the most the searches can add: ~12 cr + up to 15 cr search is three searches at 5 credits. The hold on your balance is those 15 credits plus the most text the searches, the pages and the extra model calls could add at the model's rate, on top of the usual hold for the reply. When the turn finishes the hold settles to what actually ran, and the receipt says so: 2 SEARCHES, 10 CR, with the pages read counted beside it when any were. A turn that searched but produced no answer is not charged, searches included. Holds and settlement. ## Sources and the [n] marks Under a reply that searched is a Sources line: the pages the turn read, numbered in the order it met them, each shown as its domain and opening the page in a new tab, with the title on hover. A [1] in the answer opens the same page as source 1. A page found by two searches has one number. History and Search mark a chat that searched with a small globe. ## Which models Any chat model the catalog marks as taking tools, which is most of them. The public JSON at /api/models says supportsTools per model and carries the search offer (the price and the caps); the API lists the same flag on GET /v1/models. If a provider turns out to reject the tools after all, the turn runs again without them, so you still get an answer, without sources. ## Privacy When Search the web is on, the search terms the model chooses are sent to Brave Search; page fetches are made by askr's server, never from your browser. Your account travels with neither. askr's fetch introduces itself as askr/1.0 (+https://heyaskr.ai) and honours robots.txt. Privacy. ## If it will not run Each refusal ends with a code from the list: W206 when this server has no search, W207 when the model cannot use tools, W109 when the account has started more than six searching turns in a minute, W110 when something other than the workspace asked to search without streaming. None of them charge anything. The same search runs on the API as web_search: true. Chat completions. Source: https://heyaskr.ai/docs/workspace/search --- # Attach files What can I attach to a message, which models read it, and what does it cost? Verified against the code on 2026-09-24. The paperclip in the composer gives the model a file with your message: a picture, a PDF, a text or code file, a Word, PowerPoint or Excel document, an audio file or a video. Dropping a file on the chat or pasting one does the same. Uploading is free. A file costs tokens only when a message carries it to a model, at that model's input rate. ## What you can attach | Kind | Formats | Per file | Per message | Per conversation | | --- | --- | --- | --- | --- | | Picture | PNG, JPEG, WebP, GIF | 8 MB, 8,192 pixels a side | 6 | 20 | | PDF | PDF | 32 MB | 3 | 6 | | Text and code | Plain text, Markdown, CSV, JSON, HTML, YAML and source code in most languages | 5 MB | 5 | 10 | | Document | Word (.docx), PowerPoint (.pptx), Excel (.xlsx) | 32 MB | 3 | 6 | | Audio | MP3, WAV, M4A, OGG, FLAC, WebM | 25 MB, 20 minutes | 2 | 4 | | Video | MP4, MOV, WebM | 100 MB, 10 minutes | 1 | 2 | Per conversation counts every file on every message the model is shown in a turn. What a file is comes from its bytes, not its name: a PDF called notes.txt is a PDF. An empty or damaged file is refused when you attach it, before anything is sent. ## How a file reaches the model Every model reads words. Only some read files. So when you send, each file goes to the model in one of two ways, decided per model: - As the file, where the model's catalogue entry says it takes that kind. A picture goes as a picture, a PDF as a PDF, sound as sound, a clip as a clip. - As its words, where it does not. The text of a PDF, a document or a text file, or a transcript of an audio file, is put in the message in the file's place, between lines that name the file. The model reads it like anything you typed. | Kind | Models that take the file | On every other model | | --- | --- | --- | | Picture | 212 chat models | Refused. Pick a model that sees pictures. | | PDF | 113 | Its text | | Text and code, Word, PowerPoint, Excel | None. They always go as words. | Its text | | Audio | 32 | A transcript | | Video | 66 | Refused. Pick a model that watches video. | Counted on the catalogue on 24 September 2026. So a text file, a PDF or a document can be given to any chat model. The paperclip offers every kind the model you picked can be given, which means pictures and video appear on it only for a model that takes them. The public JSON at /api/models lists each model's input_modalities; the API says the same as inputs on GET /v1/models. ## Per kind - Pictures are resized in your browser before they go: the longest side to 1,568 pixels for a chat, which is the size vision models read at, and re-encoded as JPEG. For an edit on an image model or the start frame of a video, the picture is kept at up to 4,096 pixels. Make images, Make video. - PDFs of up to 100 pages go as the file to a model that reads files: inline under 8 MB, by a link the provider fetches above it. Over 100 pages a PDF goes as its text on every model, because providers read at most 100 pages natively. A scanned PDF with no text layer has no words to send, so it works only on a model that reads files (code W611 on any other). - Text, code and documents always go as words. Word is read as its paragraphs, PowerPoint one block per slide, Excel one sheet at a time as CSV. Up to 400,000 characters a file, about 100,000 tokens; a longer file is cut there and the model is told. - Audio goes as sound to a model that hears when it is an MP3 or a WAV under 8 MB. Anything else, and any audio for a model that cannot hear, goes as a transcript, made once by a speech-to-text model and kept beside the file so a second question does not transcribe it again. If no transcript can be made, the message says so (code W608). - Video goes as the clip to a model that watches: inline under 8 MB, by a link above it. There is no words version of a video, so on any other model it is refused (code W609). ## What it costs Nothing to upload, nothing to keep. When a message carries a file to a model, the file is input tokens at that model's input rate, like the words around it. The estimate beside Send counts the files in. When you send, the hold reserves the top of what providers bill per unit, and the receipt on the reply settles to the real figure. | Sent as | Reserved | | --- | --- | | A picture | 1,600 tokens | | A PDF, as the file | 1,600 tokens a page | | Audio, as sound | 32 tokens a second | | Video | 100 tokens a second | | Words: a text file, a document, a PDF's text, a transcript | The text's own token count | A 30-page PDF on a model that reads files holds 48,000 tokens; on a model that reads its text it holds whatever the text comes to, often less. Ten seconds of video holds 1,000. Holds and settlement. ## Follow-up questions Ask about a file again and it goes again: the three newest messages of yours that carry files send them in full, so the model has them in front of it. A file on an older message is not re-sent. A one-line note in its place says a file of that name was attached there, so the model knows it existed and you do not pay for it on every reply. To put an old file back in front of the model, attach it again. Files count toward the model's context window with everything else, and a turn that does not fit says so before it runs (code W612). ## What is kept, and for how long - An upload: On your account, on askr's own storage, served to you. Nothing is public. Nobody at askr reads it, and nothing in it trains a model. - A file in a chat: Stays with the chat. Delete the chat and its files go with it; delete all your chats and every file they carried goes too. - A file in a Collab room: Belongs to the room and is readable by its members. It stays with the room, not with your chats. - An upload never sent: Swept after 7 days. - The words read from a file: The extracted text and any transcript are kept beside the file, so they are made once. - Where a file goes: Where your words go: to the model's provider, through askr's gateway, for that turn. A file over 8 MB sent as the file is fetched by the provider from a link that expires after an hour. ## The Files page Everything you have uploaded, for chats and Collab rooms, is on Files in the app: name, kind, size, what was read from it (pages, length, characters) and when. Open a file, download it, or delete it. A deleted file cannot be sent again, and a message that still names it says the attachment could not be found (code W607). ## For developers The same store is on the API as /v1/files, the OpenAI way: upload once, then name the file by file_id in a message part. A picture, a PDF, sound or a clip can also go inline in the message. Same limits, same per-model routing, same codes. Files on the API. ## If it will not go Each refusal ends with a code from the list: W604 a kind askr does not take, W605 too large or too long (the message says the limit), W606 too many of one kind, W607 a file that is gone or not yours, W608 audio for a model that cannot hear and no transcript, W609 video for a model that cannot watch, W610 an empty or damaged file, W611 no words in the file and a model that cannot read files, W612 more than the model's window holds. A picture for a model that cannot see is W203. None of them charge anything. Source: https://heyaskr.ai/docs/workspace/attachments --- # Error codes What does the code at the end of a workspace error mean? Verified against the code on 2026-09-26. Every error the workspace shows ends with a short code, like (code W513). Quote it to support and they know exactly which path failed without asking you to describe it. The groups: 1xx the request, 2xx the model picked, 3xx credits, 4xx the provider, 5xx the answer, 0xx the browser, 9xx sending credits to another @ and paying by card. In the app, the code is a chip: tap it, or hover it on a computer, for its line from this table. > Whenever the message says you were not charged, the reservation for that turn was released before the message was written, and the wallet shows nothing for it. | Code | What you see | What it means | What to do | | --- | --- | --- | --- | | W101 | Please sign in to use the workspace. | The session has ended or was never started. | Sign in again. | | W102 | That run could not be found. | A reloaded page asked for an answer that finished more than fifteen minutes ago. | Send the message again. | | W103 | Model access is switched off on this server right now. | The server has no gateway key. This should never show on heyaskr.ai. | Tell support. | | W104 | Type a message first. | The request carried no text and no image. | Type something. | | W105 | That request carries a message type the workspace does not accept. | Something other than the workspace sent a system or tool message. | Use the workspace, or the API for system prompts. | | W106 | That message is too long for this model. | One message is over what the model's window allows: 48,000 characters in the workspace, or 24,000 in a collab, on a model with a small window. | Pick a model with a larger window, such as Kimi K3 at 1 million tokens, or split it. | | W107 | That message is too long to send even on its own. | The newest message alone exceeds what one turn can carry. | Start a new chat with the part that matters. | | W108 | At most 20 images in one conversation. | The conversation already carries the image limit. | Start a new chat to send more. | | W109 | Search the web is limited to a few turns a minute. | More than six searching turns were started in a minute on this account. | Wait a moment and send again. | | W110 | Search the web only works on a streamed chat turn. | Something other than the workspace sent search on without streaming. | Send it with stream on. | | W201 | That model is not in the catalogue right now. | The picked id is no longer listed by the gateway. | Pick another model. | | W202 | That model does not answer chat. | A video or audio model was sent to chat. | Pick a chat model, or use the right door. | | W203 | That model cannot see pictures. | An image was attached for a text-only model. | Pick a model marked for image input, or remove the picture. | | W204 | That model's provider is not serving it at the moment. | The model is on our known-dead list. You were not charged. | Pick another model. | | W205 | That model has no price yet. | The gateway lists it with no rate, so it cannot be billed. | Pick another model. | | W206 | Search the web is not available on this server right now. | The server has no search key. You were not charged. | Turn Search the web off and send again. | | W207 | That model cannot search the web. | The model does not take tools, or paints pictures. You were not charged. | Turn Search the web off, or pick a model that can. | | W301 | Not enough credits for this turn. | The reservation for this turn is more than the balance: yours, or the room's in a collab that pays from the room balance. | Top up, or pick a cheaper model. In a collab that pays from the room balance, add credits to the room: any member can. | | W302 | This account is paused. | A card payment was reversed after its credits had been spent, and those credits are owed. Runs and sends stay paused while anything is owed. | A crypto top-up pays what is owed first; a card payment waits until nothing is owed. Contact support if you think this is wrong. | | W411 | We could not reach the model provider. | Our gateway did not answer. You were not charged. | Try again in a moment. If it persists, support checks the gateway. | | W412 | The model provider turned this request down. | The gateway returned an error that is not a content refusal, a limit or an outage. You were not charged. | Try again, or pick another model. Support can see the status in the logs. | | W413 | The model provider is busy. | The gateway rate-limited us. You were not charged. | Try again in a few seconds. | | W414 | Model access is paused for a moment on our side. | Our gateway account needs topping up. Nobody is charged. Also shown on the Video door. | Support tops up the gateway. Try again shortly. | | W415 | The model's provider declined this under its content rules. | The provider refused the request on content grounds, before any answer. Asking for a real public figure is a common cause. You were not charged. | Reword it, or pick another model. Rules differ by provider, so another model may accept it. Trying again unchanged will be refused again. | | W416 | The (direct) version is paused for this account for 24 hours. | A maker's own API, Claude (direct) or GPT (direct), declined too many of this account's requests under its content rules within a day, so that maker's (direct) versions are paused for it. Nothing was charged. Every other model, the same model's version without (direct) included, still runs. | Wait a day, or use another model. If the refusals were mistaken, contact support. | | W511 | The connection to the model dropped before it answered. | The stream died before the first token. You were not charged. | Try again. | | W512 | The model answered in a form we could not read. | A 200 with an unreadable body. It was billed upstream, so the reservation was charged. | Tell support the code; the turn is refunded by hand. | | W513 | That model did not answer this time. | The provider finished with no content. You were not charged. | Try again, or pick another model. | | W514 | That model spent its whole output budget thinking. | A reasoning model used its whole output budget (up to 32,768 tokens) on thinking. You were not charged. | Try again, shorten the prompt, or pick a model that thinks less. | | W515 | The model's provider reported an error. | An error arrived inside the stream. You were not charged. | Try again, or pick another model. Support can see the provider's text in the logs. | | W001 | The connection dropped before the reply arrived. | The browser lost its connection to heyaskr.ai mid-turn. | Check the connection and try again. | | W002 | The model did not return a response. | The server's answer carried neither text nor an error. | Try again. If it repeats, tell support. | | W601 | Images must be sent as a data URL. | The attachment reached the server in a shape it does not accept. | Attach the picture again from the clip button. | | W602 | Images must be PNG, JPEG, WebP or GIF. | Another format was attached. | Convert it, or screenshot it. | | W603 | An image part needs image_url.url. | An API-style message carried an empty image part. | Use the workspace to attach pictures. | | W604 | That file type is not accepted. | The upload is not a picture, PDF, text or code file, Word, PowerPoint or Excel document, audio or video, judged from its bytes rather than its name. | Convert it, or paste its text. | | W605 | That file is too large. | Over the limit for its kind: 8 MB for a picture, 32 MB for a PDF or document, 5 MB for text, 25 MB or 20 minutes for audio, 100 MB or 10 minutes for video. The message says which. | Shrink it, split it, or trim it. | | W606 | Too many attachments of one kind on a message. | More than 6 pictures, 3 PDFs, 5 text files, 3 documents, 2 audio files or 1 video on one message, or more than the per-conversation allowance in what is sent. | Send them across several messages, or start a new chat. | | W607 | That attachment could not be found. | The id names nothing on your account: deleted, swept after seven days unused, or someone else's. | Attach the file again. | | W608 | That model cannot hear audio and no transcript could be made. | The model takes no audio and the speech-to-text step failed or is not offered here. | Pick a model that takes audio. | | W609 | That model cannot watch video. | A video was attached for a model whose catalogue entry takes no video. | Pick one that does, or remove the video. | | W610 | That file could not be read. | Empty, damaged, or not what it claimed to be. | Open it locally to check, then attach it again. | | W611 | No words could be read from that file, and this model cannot read the file itself. | A scanned PDF with no text layer, or an empty document, sent to a model that takes no files. | Pick a model that takes files, or attach the text. | | W612 | That message, with its attachments, is longer than this model can take in. | The words plus the files come to more than the model's context window. | Pick a model with a bigger window, or attach less. | | W701 | That image could not be found. | Generated images are kept for 24 hours. | Make it again. | | W801 | That render failed. | The video provider gave up on the clip. You were not charged. | Try again, or pick another model. | | W802 | That video could not be found. | The clip id is unknown or belongs to another account. | Open it from your library. | | W803 | That video is no longer available. | Older clips expire; newer renders are kept on your account. | Render it again. | | W804 | Describe the video you want. | The prompt was empty. | Type a description. | | W805 | That description is too long. | Video prompts have a short limit. | Shorten it. | | W806 | That is not a video model. | A chat model was sent to the Video door. | Pick a video model. | | W807 | That model does not take a start frame. | A picture was attached for a text-to-video model. | Remove the image, or pick an image-to-video model. | | W808 | That model needs a start frame. | An image-to-video model was sent text only. | Attach a picture. | | W809 | No published price for this length and shape. | The gateway prices this model per shape, and this one is missing. | Pick another length or aspect ratio. | | W810 | Not enough credits for this video. | The clip's price is more than the balance. | Top up, or pick a shorter clip. | | W811 | The video provider would not start that render. | The gateway refused the job. You were not charged. | Try again in a moment. | | W812 | That render is already in progress. | A retry was sent while the same clip was rendering. | Wait for it. | | W813 | That render could not be found. | A reloaded page asked about a render the server no longer has. | Send it again. | | W815 | That model cannot start from a picture. | Its listing says it takes a start frame, but its provider refused one. You were not charged, and the frame is no longer offered for that model. | Render from the prompt alone, or pick another model. | | W816 | The video provider declined this render under its content rules. | The provider refused the prompt or the start frame on content grounds, when the render was sent or while it ran. A real public figure is a common cause. You were not charged. | Reword the prompt, or pick another model. Rules differ by provider, so another model may accept it. Trying again unchanged will be refused again. | | W817 | The video provider is not taking renders right now. | The provider was down or had no capacity for that model. You were not charged. | Try again in a few minutes, or pick another model. | | W818 | The video provider is busy. | The provider rate-limited the render. You were not charged. | Try again in a minute. | | W819 | The video provider took too long and gave up on that render. | The render timed out on the provider's side. You were not charged. | Try again, or pick another model. | | W820 | Type what to say, or describe the music or the sound, first. | The request carried no text. | Type something. | | W821 | That is too long for this model. | A reading is over what the voice takes (5,000 characters, 3,000 on ElevenLabs v3), or a description is over 2,000. | Shorten it, split it, or pick a model that takes more. | | W822 | That audio model is not available right now. | The id is not in the audio catalogue, or its supplier is switched off. | Pick another model. | | W823 | That voice is not one this model has. | The voice name is not on the list the door offers. | Pick a voice from the list. | | W824 | That length is outside what this model makes. | The length asked for is outside the model's range, or the request could not be priced. | Pick a length in the model's range. | | W825 | Not enough credits for this audio. | The balance cannot cover the quote: yours, or the room's in a collab that pays from the room balance. The message says how much is needed. | Top up, or pick a cheaper model. In a collab that pays from the room balance, add credits to the room: any member can. | | W826 | That audio could not be found. | No audio with that id on this account, or it was deleted. | Make it again. | | W827 | The audio provider would not start that. | The supplier turned the job away before it began (down, busy, or our account with it). Nothing was reserved. | Try again in a moment, or pick another model. | | W828 | The audio provider refused that under its content rules. | The supplier's filters stopped it: a real artist's voice, copyrighted lyrics, or content it will not make. You were not charged. | Reword it, or pick another model. | | W829 | That audio failed on the provider's side. | The job was taken and then failed or timed out. The hold was released. | Try again, or pick another model. | | W901 | Send at least 100 credits. | The amount is under the minimum, or is not a whole number of credits. | Enter 100 or more, whole credits. | | W903 | Credits deposited in the last 24 hours can be sent tomorrow. | The amount is within your balance, but part of the balance arrived by deposit in the last 24 hours and is still cooling off. | Send what Sendable now shows, or wait. | | W904 | No one on askr has that @. Check the spelling. | No active account has that @, or the field does not hold one (3 to 32 letters, digits or underscores). | Check the @ with them. An account without an @ picks one on its profile. | | W905 | You cannot send credits to yourself. | The @ is your own. | Pick someone else's @. | | W906 | Enter the six-digit code we emailed you. | The send is above 5,000 credits, so it needs the code; one was mailed to the account's address when this showed. | Enter it in the dialog. Press Send a new code if it does not arrive. | | W907 | That code is not right, or it has expired. Ask for a new one. | The code was wrong, already used, or older than ten minutes. Five wrong guesses burn a code. | Press Send a new code and enter the new one. | | W908 | That is more than you can send right now. | The amount is over what is sendable now: balance minus open holds minus deposits in the last 24 hours. Nothing was sent. | Wait for a running request to settle, top up, or send less. | | W912 | Credits bought by card cannot be sent to other people. | Your balance covers the amount, but only by counting credits bought by card, which never move. The message says how many credits can be sent now. Nothing was sent. | Send what Sendable now shows. Pay by card. | | W913 | Sending is paused on this account while a card payment is disputed. | A card payment on the account is in dispute, so sending credits, funding a room and a new card payment all wait. The pause lifts when the dispute closes; if it is lost and leaves credits owed, the account is paused under W302 instead. Nothing was sent, charged or added. | Contact support if you think this is wrong. | | W914 | Tick the box to say you want your credits now. | A card checkout opens only once you have said, in askr, that you want the credits on your balance straight away and understand a card payment cannot be cancelled once they are there. The box was not ticked when the checkout was asked for. Nothing was charged. | Tick the box above the payment buttons, then choose how to pay. | Source: https://heyaskr.ai/docs/workspace/error-codes --- # Make images Which models make pictures in the workspace, and what do they cost? Verified against the code on 2026-09-24. Some chat models answer with a picture. Pick one of those in the picker, describe what you want, and the image appears in the conversation. There is no separate image tool to learn. The models that do this are marked in the catalog as chat models whose output includes images: today the three Gemini image models (Nano Banana). ## Cost A picture is priced by what it actually costs, not by the per-token rate on the card, because most of the tokens are the picture. The picker shows the measured price for the models we have measured (41 to 163 credits) and reserves 200 for any we have not; the charge settles to the real figure. What you are charged when. ## Keeping it Every picture you make in the workspace is saved to your library with its prompt, model and cost. Download it from there any time. ## Editing and image input Attach a picture from the paperclip and ask for the change you want: a different background, a new colour, the same scene at night, a variation. The image models that take a picture in, the Gemini and GPT image families, 9 models on the catalogue on 24 September 2026, make a new picture from yours. For an edit the picture goes at up to 4,096 pixels on its longest side, not the 1,568 a chat uses, so the result keeps the detail of the original. The new picture is priced like any other, and it is saved to your library beside the prompt. A picture attached to a chat model that sees pictures is read, not redrawn: ask what is in it, what a screenshot says, or what a chart shows. Attach files says which models see pictures and what one costs. Source: https://heyaskr.ai/docs/workspace/images --- # Make video How do I make a clip, what does it cost, and how long does it take? Verified against the code on 2026-09-24. Video runs in the workspace only. Pick a video model, describe the clip, choose a shape and a length, and the exact price is shown before you start. The API does not have a video endpoint yet. 1. Choose a video model. Kling, Veo, Runway, Seedance, Hailuo and others. The catalog lists them with their per-clip prices. 2. Choose shape and length. Each model renders a fixed set of aspect ratios (16:9, 9:16, 1:1) and durations (typically 5, 8 or 10 seconds). The picker only offers combinations the model publishes a price for. 3. See the price, then render. The price is the model's published price for that exact shape and length, in credits. Nothing is estimated. 4. Wait. Renders take from under a minute to several minutes. The clip shows its first frame while it works, and you can leave the page. ## Starting from a picture A model that takes a start frame takes a picture from the paperclip. It goes at full size, up to 4,096 pixels on its longest side and 6 MB, so the clip starts from the detail of the original rather than a copy shrunk for chat. A model that cannot start from a picture says so before the render (code W815), and one that needs a picture asks for one (code W808). Attach files. ## Billing - The price is reserved when the render starts and held for up to 20 minutes. - A render that fails releases the hold. You are charged only for a clip that completes. - Ten renders per minute per account. - The description is limited to 2,000 characters. ## Keeping the clip Finished clips are saved to your library and can be downloaded as MP4. They are served from askr, to you only. ## Errors you may see | Message | Meaning | | --- | --- | | That model renders X, Y second clips. Ask for one of those. | The length you asked for is not one the model offers. | | That model has no published price for this length and shape. | The combination cannot be billed, so it is refused before anything runs. | | Not enough credits for this video. | The balance cannot cover the price. The message says how much is needed. | | The gateway would not start that render. Nothing was charged. | The provider refused. Try again or change the prompt. | | That render is already in progress. | The same request was sent twice. | Source: https://heyaskr.ai/docs/workspace/video --- # Make audio How do I make a voice, a song or a sound, and what does it cost? Verified against the code on 2026-09-25. The Audio door makes three things: a voice reading your text, a song from a description, and a sound effect from a description. Pick what to make, pick a model, and the exact price is shown on the button before you start. Audio runs in the workspace only; the API has no audio endpoint. 1. Choose what to make. Voice, Music or Sound, under the composer. Each has its own models. 2. Choose a model, and a voice. A reading takes one of twelve ElevenLabs voices, each described in the list. A song takes a description of genre, mood, tempo and what it sings about; a sound takes a description and a length. 3. See the price, then send. A reading is priced by its length in characters, a song per song, timed music per minute, a sound per second. The number on the button is the price. 4. Play it. A reading lands in a few seconds, a sound in about ten, a song in under a minute. The card plays it, shows the lyrics a song sings, and downloads it as MP3. ## Models | Make | Model | Priced | | --- | --- | --- | | Voice | ElevenLabs Turbo 2.5 by fal, ElevenLabs v3 by fal, ElevenLabs Flash 2.5 | Per 1,000 characters | | Music | Lyria 3.5 by fal, Lyria 3 Pro | Per song, whatever its length | | Music | ElevenLabs Music by fal | Per minute, at the length you pick | | Sound | ElevenLabs Sound Effects by fal | Per second, at the length you pick | A model labelled "by fal" runs at fal, the supplier that carries ElevenLabs and Google's Lyria for us; the others run at our default gateway. Lyria writes whole songs with vocals, structure and lyrics, up to a few minutes. ElevenLabs v3 takes cues in square brackets, like [laughs] or [whispers]. ## Billing - The price is reserved when the job starts and held for up to 20 minutes. - You are never charged above the price shown. A job that fails releases the hold. - If the supplier will not start a job, the same request runs at the other supplier's model at that model's price, and the receipt says which ran. - Twenty audio requests a minute per account. A reading takes up to 5,000 characters (3,000 on ElevenLabs v3); a description up to 2,000. ## Keeping it Finished audio is saved to your library with the prompt that made it, the voice and the lyrics, and can be downloaded as MP3. It is served from askr, to you only. Delete any of it from the library. ## Made with AI Every reading, song and sound carries a line saying it was made with AI, and a reading says it is a synthetic voice. Under the EU's AI rules in force since August 2026, synthetic audio has to be recognisable as such; keep the line when you pass it on, and do not present a synthetic voice as a real person's. The door offers premade voices only: it does not clone a voice from a recording. ## Errors you may see | Code | Meaning | | --- | --- | | W821 | The text is over what the model takes. Shorten or split it. | | W824 | The length is outside the model's range. | | W825 | Not enough credits for the price shown. | | W827 | The supplier would not start the job; nothing was charged. | | W828 | The supplier refused it under its content rules: a real artist's voice, copyrighted lyrics, or content it will not make. | | W829 | The job failed after it started; the hold was released. | Source: https://heyaskr.ai/docs/workspace/audio --- # askr Mind What does askr remember between models and chats, where does it go, and how do I control it? Verified against the code on 2026-09-26. askr Mind is the memory every model on askr shares. What you told one model is there for the next: switch from a coder to a writer to an image model and nobody has to be told again who you are or what you are building. It is one short block per chat and per room, and one about you, each under your control. ## What it remembers - Who is here. The people in the room, by @, and what each is doing. - Building. What the chat or the room is making. - Decided. The choices made, one line each. A decision changed later replaces its old line instead of sitting beside it. - Open. Questions still open. A settled one leaves the list. - Made. Every picture, clip and sound made in the chat or the room, with the model that made it and a caption. Written by askr's own code from what was made, never by a model. - Preferences. How each person likes things: any "I like", "never", "always" or "I want" a person says, kept with their @. Each entry is one short plain sentence. After every answer a fast, cheap model reads the lines since it last looked and adds entries or replaces one; it never rewrites the whole memory and never deletes a line, so nothing you saw there disappears on its own. Nothing anyone types is an instruction to it. ## Where it goes The block goes first into every prompt, before the conversation, so it is the first thing every model reads. Every model in a room reads the same block: the one you switch to knows what the one before it knew. While a model is being briefed the chat says Using askr Mind until its first word; in a room the same words stand on the busy line, and before a picture, a clip or a sound is made. The Building and Made lines are also what askr Pro reads before it picks a door and a model for a line (askr Pro). A room's block is titled Room memory, a chat's Chat memory, a project's Project memory, and the one about you About the person. It stays short whatever a room accumulates: at most 12 people, 3 Building lines, 25 Decided, 8 Open, 30 Made and 12 Preferences, about 1,500 tokens in all. Past a cap the newest Made and Who is here lines are kept and the oldest Decided lines are, since a decision made early still holds. ## What you can do - Pin. Beside Reply on any line in a room, Pin puts that line's words into the memory, under the heading that fits (askr picks it, Decided when unsure). A pinned line is marked as yours and a toast says "Pinned to askr Mind". - Edit. In the Memory card (Room settings in a room, Settings, Memory for your account) click an entry, change the words, Enter saves. An edited entry counts as yours. An entry is at most 600 characters. - Forget. Forget beside an entry removes it, and askr Mind is told never to add those words again. - Forget everything. Clears the whole memory of that chat, room, project or account, after a second press to confirm. - Switch it off. Remember this room in Room settings, the creator's switch; Remember me across chats under Settings, Memory. Off, nothing new is remembered and no model is told what is there; what is there stays until you forget it. - A temporary chat. The dashed-bubble button in the chat's message box. On, the chat says "askr Mind is not reading or writing this chat": no block goes in, nothing is written, and a picture made there is not noted. Remembered per chat. - Mind noted. After an answer a chip under it says what was just remembered, "Mind noted: ..." or "Mind noted 3 things"; tap it to open the memory. It goes when the next line arrives. ## The Made list, and Prompt used Every picture, clip and sound made in a chat or a room is noted under Made the moment it is shown, with the model and the request. Once a picture is on screen a vision model writes one line on what is in it, its style and its mood, and that caption takes the request's place in the entry. A clip or a sound keeps its request as its caption for now. When you ask the Image, Video or Audio door for something, askr Mind writes the prompt the generator gets from the memory and your request: what the room is building, the look it decided on, the picture you mean by "that one". Under the result a folded row, Prompt used, opens to the exact words the generator received, with a thumbnail of the picture it built on when there was one. A request that stands on its own goes as it is, and so does any request when the writer takes more than twenty seconds. A made picture starts a clip. Ask the Video door for the picture, moving, and it goes to the model as the start frame, on a model that takes one; ask the Image door for a change to it and the picture goes beside the prompt to a model that takes a picture in. Only the first picture a request refers to is handed over. Make video, Make images. ## Recall The memory is short on purpose, so the rest of the room stays reachable by search. When a thread holds more lines than the prompt carries, the words of your request are searched across every line of the room, and up to five earlier lines that match go to the model as "Earlier in this room, found by search", each with who said it. The search is by word, in any language, and never reads a line that came after the ones in the prompt. ## About you, across chats Beside each chat's and room's memory, askr Mind keeps one list about you across chats: who you are, what you build, how you like things. It learns at most three lines from an answer, never the details of one task, and at most once every five minutes. It goes in front of every chat of yours, before the chat's own block. Settings, Memory shows it, with the switch Remember me across chats, the entries to edit or forget, and Forget everything. ## Projects A project is one memory shared by the rooms and chats you put in it. Make one under Settings, Memory (up to 100), name it, and pick it beside a chat's title or in Room settings. Its memory goes into every prompt of those rooms and chats, after the block about you and before the room's or the chat's own, and what is said in any of them folds into it, at most once every five minutes. Delete a project and its rooms and chats stay, out of it, and its memory is forgotten. - Shared: The project also brings what askr knows about you into its rooms, under "About @you", whether or not you switched that on in the room. The default. - Walled off: The project's memory goes in and nothing about you does. ## Team projects A project can hold a team, so rooms in it know what the other rooms are doing, across people. Under Members on the project's row the owner invites people by @; the invited person sees "@name invited you to " at the top of their Projects card and accepts or declines. What is shared is each room's memory lines, its search hits and the things it made, which reach the other rooms of the project, and the project's memory carries a Rooms heading, one line per room on what it is doing now, each linked to its room. The room's transcript itself is never opened to the others. A room goes into a project by its creator's pick in Room settings, from the projects they own or were let into; a public room never joins one. The owner edits the project's memory and its sharing; members read it, and their rooms take part. Leave a project and your rooms come out of it with you; the owner can remove a member, with the same effect. ## Rooms with other people A room's memory belongs to the room: everyone in it reads the same block, any member can pin, edit and forget, and the creator holds the Remember this room switch. What askr knows about you stays out of a room unless you switch on Bring my memory into this room in Room settings, your own choice for that room only. Once it is in, the room's models read it as "About @you", so anyone in the room can ask them about it; nobody sees the list itself. A public room never takes it: the switch is refused with "Your memory stays out of a public room." askr Collab. ## What it costs Nothing, today. askr Mind's own calls, the fold after each answer, the caption, the prompt writer, are metered like everything else and charged at zero credits while askr is feeless. The block itself is part of the prompt, so it counts as input tokens on the model you run, like anything else in the conversation; the hold before a turn counts it in. None of it holds a turn up: the fold runs after the answer is shown, the caption after the picture, and a prompt writer that fails or runs late is skipped. A failure in askr Mind never fails a turn. ## Privacy and deletion The memory is stored on askr beside the chat or the room it belongs to, and reaches a model's provider only inside a prompt, the way the conversation does. Nobody at askr reads it and nothing in it trains a model. Delete a chat and its memory goes with it; close and delete a room and the room's goes; delete a project and its memory is forgotten; delete your account and all of it goes. A chat's export and a room's Export carry the entries. Privacy, Chat. ## If it will not run askr Mind carries no error codes of its own; its refusals are plain words. The Memory card says "askr Mind is not answering right now. Try again in a moment." when the memory could not be read, and "askr Mind is switched off on this server." when the whole feature is off. An edit of a blank line is refused with "Say what the entry should read.", one over 600 characters with "An entry is at most 600 characters.", and an entry that is gone with "No such entry." A project is "No such project." when it is not yours, and "You have 100 projects already. Delete one first." at the cap. Nothing here charges anything. Source: https://heyaskr.ai/docs/workspace/mind --- # askr Pro What is askr Pro, what does it read before it picks a door and a model for my message, and what are its rules? Verified against the code on 2026-09-27. askr Pro is a model you never have to pick. A new chat opens on it, at Balanced, and it sits at the top of the model picker, the first row, with Cheap, Balanced and Performance beside it. Say what you want and askr Pro reads the message and picks the door and the model for it, at the tier you chose, starts one of the ten skills when the line calls for one, and says what it picked under your line. Version one picks from one ladder per door and kind of task; version two builds the ladder around you. ## On from the start A new chat opens on askr Pro, Balanced, so there is nothing to turn on: every line you send is decided, which of the room's doors, Ask, Code, Image, Video or Audio, and which model behind it. With askr Pro on there are no door chips under the message box, since it picks the door on every line, and no estimate beside Send, since the price is not known before the pick; a model picked by hand brings both back. The Images, Video and Audio doors on the home page still start a room on that door. Picking by hand still works and sticks: choose a model from the picker and the chat runs on it from then on, open a skill from Ask and that thread runs under the skill, until you choose askr Pro again from the first row of the picker, with a tier on it. In a room the creator's pick sets askr Pro for the room; anyone else's holds for their own lines. On the home page the Ask composer opens on askr Pro too, and the first send makes a room on it; the tier, and a model you picked by hand instead, are remembered on that device, so the next new chat opens the way you left it. - Cheap: Always the cheap ladder: a cheaper model and short answers, however hard the question. A hard one moves up to the tier's own reasoning list, never out of the tier. - Balanced: Strong models at an everyday price. A hard question goes to the balanced reasoning list. - Performance: The strongest models, thorough answers and more searching, and always the best there is for a picture, a clip or a sound. A hard question goes to the strongest reasoners. ## What it reads before each pick - The message, and what is attached to it. A picture, a PDF, a sound or a clip on the line changes what the line is likely to want. - The last lines of the thread, so a follow-up is read as a follow-up and not as a new task. - What the room remembers: the Building and Made lines of the room's memory, so it knows what you are making and what has already been made. askr Mind. - The tier, and the doors the room has on. A door the creator switched off is never picked. ## How it decides That reading goes, in one call, to a decision model that answers with probabilities, not words: how likely each door is, which of thirteen kinds of task the line is (a quick answer, writing, code, working out, looking it up, a summary, translating, money, health, ideas, shopping, fixing it, learning), whether it wants the web, how hard it is, whether it carries on the last one, and whether it also wants a picture, a clip or a sound made beside the words. Plain code then turns those numbers into a door and a model from the tier's ladder. The decision model never writes a word, so it cannot answer you, cannot change your request, and nothing you type is an instruction to it; the tier is held by the code, not by a model, so a message that asks for a dearer model does not get one. How many pictures you want is read from your words by a plain pattern, and a clip's length from the room's setting, never from the decision model. The decision is the only thing that runs before the credits for the line are held, and it has two and a half seconds. When the decision model is slow or down the line still runs, on the room's model through Ask, and the line under it says askr Pro could not decide. Each decision is kept as a row with no message text in it, so askr can see how often it decides and how sure it was. ## The rules - The tier is a ceiling and a floor. Cheap always runs the cheap ladder, Balanced the balanced one, Performance the top one, and Performance always picks the best there is. A hard question moves up to the tier's reasoning list, never out of the tier, and nothing in a message moves it. - A follow-up keeps the model. When a line carries on the last task and the model that answered it is still live and inside the tier, it answers again, and the pick says kept. A new task picks afresh. - A picture, a clip or a sound needs a confident read. Under that bar the line goes through Ask, with a chip under it, Make it a picture instead (or a clip, or a sound), that runs the same line through that door. Ask and Code need no bar: they are cheap either way. - A skill, without asking. A line that calls for one of the ten skills, looking something up, writing, learning, fixing something, health, translating, money, ideas, a summary or shopping, runs with that skill's instructions and its screen: the answer arrives with the sources, steps, cards or calculators under it, in the same thread, and the pick says which skill ran. A thread that already carries a skill keeps its own. askr Skills. - Every pick is shown. Under your line, one line says what ran and why, askr Pro: Claude Sonnet 5 for writing, Balanced, or askr Pro: Look things up skill on Claude Sonnet 5, Balanced, with Change beside it. - Change turns askr Pro off. Pick a model by hand from Change and the line runs again on it; askr Pro is then off for the room, and the model you picked stays as the room's model, so the change sticks. Choosing any other model in the picker does the same. - A door you ask for is honoured. Make it a picture instead runs the line through the Image door, and a line sent to a named door gets that door; askr Pro then picks only the painter, the renderer or the voice for the tier. ## Made beside the words When a line wants a picture, a clip or a sound as well as words or code, a site and its branding, a plan and its poster, the model that answers makes them itself as it writes. It writes each prompt, each one runs through the room's own door on the tier's painter, renderer or voice, every made thing lands under the answer as its own line, and the room says making a picture while it does. It is told to make things itself and never to send you to a door. Code it writes carries the pictures it was made with, by their addresses, so a page it builds shows them without a file to drop in. ## Packaged, previewed and on the bill After a run that made things askr Pro checks once, off the turn, whether everything you asked for is there and connected; whatever is missing is made in one more run, never more, and a closing card under the last reply sums the package up in a line, then the bill for the whole line in credits and about how many dollars, split by clips, pictures, sounds and code or words, a clip still rendering counted when it lands. A failed line is left out; its credits went back. When a thread has a page to show, the preview takes the room: on a desktop the rail folds away and the panel takes the right half at full height, Preview first, with a small button that reaches Room settings and People; on a phone one switch moves between the chat and the full-screen preview, with Chat to come back. The page is assembled from the answer's files, its stylesheets and scripts inlined by name, and every picture, clip and sound made in the thread is put into the page before it is shown. Anchors scroll, a link to another page of the answer shows that page with Back, a link to a made picture, clip or sound opens it in the room, an outside link opens a new tab, and the frame never leaves the page. A Files tab beside Preview and Code lists everything the thread made, pictures, clips, sounds and code files, each with Open and Save; a Made here card in the rail, and at the top of the Room sheet on a phone, shows the newest six pictures and clips with All opening the Library. In every answer the fences fold into one row, Your code, 3 files, closed until you open it, each file with Copy and Save; a short snippet with no file name stays inline. askr Collab, Make images. ## Never stopped in your tracks When a supplier declines a picture, a clip or a sound under its content rules, askr Pro rewords the request once, the brand, company, real person, franchise or copyrighted work replaced by the kind of thing it is, says so in the room, and tries again. A second refusal becomes an offer on your line: what could not be done and what would be done instead, with a chip, Yes, make the sound that way, that runs it through the door. Inside a written answer the model rewords and tries again itself. With askr Pro off a refusal reads as it always did, with its code (error codes). ## What it costs A fraction of a credit per decision, metered under its own source and charged at zero while askr is feeless. The model that answers is priced as always, and its receipt says which model ran and which pot paid. There is no estimate beside Send while askr Pro is on and nothing asks you before a dear pick, since the price is not known before the pick: the credits show under the answer, as they do on every answer, and a decision that could not be made costs nothing. What a run costs. ## Where it is not A skill thread you opened yourself keeps its tier pill: the tier there already picks the model, so askr Pro is not offered in it; a skill askr Pro starts runs inside your plain thread, where askr Pro stays on. Compare names its own models, two to four of them, and never takes askr Pro. The API names a model on every request and does not take askr Pro. askr Skills, Compare, Chat completions. ## What is next Version two builds the ladder around you. From what your memory says you reach for, what you ask and what you keep, askr Pro will rank the models per person, so the one it picks for you is the one that does your kind of work best, whether that means the strongest answer or the cheapest one. Roadmap. Source: https://heyaskr.ai/docs/workspace/pro --- # askr Collab How do rooms work: who pays, who can join, what a room can do, and what its errors mean? Verified against the code on 2026-09-26. askr Collab is a room where people work the same models together, out of one purse, with everyone seeing every answer as it arrives. Ask, Code, Image, Video and Audio, the same doors as the workspace. A new chat of yours is a room too, with one person in it, so everything below applies to your own chats as well: threads, Re-run, Compare, Pin and Export. ## Rooms and invites - Starting one: From the Collab tile, or New room: a title, a model, and who can see it. Invite only is the default. - Inviting: By @. The invited person sees the room in their list and accepts or declines. The room is not listed and not searchable; the link is the door. - Members: Up to 12 in a private room. Anyone can leave; the creator can remove a member, and a member who leaves takes back the unspent part of what they funded. - Closing: The creator closes a room once nothing is running in it. Whatever the purse still holds goes back to the people who funded it, in their shares. A closed room can be read, not written; delete it after that and its pictures, clips, sounds and files go too. ## The purse, and who pays A room with people in it pays from its own balance, the purse. The creator adds credits from their own balance, up to 100,000 credits in a room, and every run in the room is paid by the room at the model's price. Nobody spends their own balance by being invited, and a viewer of a public room never pays. Credits bought by card can fund a room; whatever comes back to you comes back as card credits. A room of one, nobody invited, pays from your balance exactly as a chat does: free credits first, then credits you were given, then the credits you bought, with the holder discount on the bought part, and the receipt says which. A run started alone settles alone, even if you invite someone while it runs. Once someone is in the room with you, the room pays, so it needs credits added first. Credits and pots, Holds and settlement. ## Talking, and asking - @askr. With people in the room, a line runs only when it says @askr; any other line is talk, seen by the room and by the model as part of the conversation. In a room of one every line runs. - One name, several models. @askr is whichever model the room, or the line, runs on. The model is told which lines it wrote and which another model did, and which door made a picture, a clip or a sound. - A model per line. The composer's model picker sets the model for the next line without changing the room's default. The answer says which model wrote it. Pick askr Pro, the picker's first row, and the room picks the door and the model for each line itself (askr Pro). - Doors. Ask and Code are the room's text doors; Image, Video and Audio pills under the message box switch the line to that door, at the model and price shown. The model is told which doors the room has and to say which one makes what you asked for, since it cannot use one itself. - Search the web. The globe in the room's composer, per room; a line that needs live facts searches and its answer carries its sources, paid by the room. Search the web. - Files. The paperclip attaches pictures, PDFs, text and code, documents, audio and video to a line; the file belongs to the room and stays with it. Attach files. ## On every line - Reply quotes the line under yours. React adds an emoji, up to 24 reactions on a line. Copy takes the words, the picture or the clip's link. - Edit your own line; everyone sees it marked edited. - Pin puts the line's words into the room's memory. askr Mind. - Re-run, on the newest answer, asks the same question again on the model picked in the composer. The old answer stays; the new one answers the same line and reads the thread only up to it, so its earlier answer is not in the prompt. A picture line runs through the Image door once; a clip or a sound cannot be re-run. - New thread continues from that line in a thread of its own, below. - Receipts. Every answer shows the model, how long it took, the credits, and which pot paid. ## Threads and branches A room starts with one thread, General. Make more from the Room card, up to 50, each with its own history; the memory is the room's, shared by all of them. New thread on any line makes a thread that continues from there: it reads the room up to that line, then its own lines, so you can take a conversation another way or edit what you asked without losing the original. A thread made from a skill thread keeps the skill and its settings. askr Skills. ## Public rooms On the People card, or when you start a room, choose Anyone with the link. Then anyone with the link watches the room live, its pictures, clips and sounds included, with no account; anyone with an account and an @ can Join, up to 100 people, and past that the link is view-only. Invite only puts it back the way it was; whoever joined stays. The purse pays in a public room too, and a viewer never does. A shared link unfurls with the room's title; a private room's card names none. What askr knows about you never enters a public room. An account made from a public room you shared counts as your referral. Referrals. ## What a room can do Room settings has a Can use row: the creator turns Ask, Code, Image, Video and Audio on or off for the room, and the composer offers only what is on; a line sent to a door that is off does not run. Every member, and every viewer of a public room, sees the settings and the room balance; only the creator changes them, adds credits, or closes the room. Room settings also holds the room's project and the askr Mind card. ## Compare In Ask and Code, Compare beside the model picker opens the picker as a multi-select: pick two to four models and the line goes to all of them, side by side, each answer streaming in its own cell with its own receipt. Every run holds its credits before any of them starts, so nobody is charged for two of four: when one cannot start, none does, and the room says why. Pick marks the answer the room keeps: the others fold to a head row (Show opens one again), the composer follows the picked model, and later lines and the room's memory read the picked answer only, or the first one while nothing is picked. Stop stops all of them. Compare runs on the Ask and Code doors, and a compare never hands a line to another model when one of them runs out of room; that one fails as any other. ## Export Export in Room settings saves everything in the room as one JSON file: the room and its settings, every thread, every line with its files and receipts, and the memory. Members only, and always the live copy, never a cached one. ## The memory card Room settings holds the askr Mind card: how many things the room remembers, the entries by heading, click to edit, Forget, Forget everything, and the creator's Remember this room switch. Bring my memory into this room is each member's own. Every model in the room reads the same memory, and what was made in the room is listed there with a caption. askr Mind. A room can sit in a team project: the creator picks it under the Memory card, from the projects they own or were let into, and a public room never joins one. Then the room shares what it does with the project's other rooms, across the people in the project: its memory lines, its search hits and the things it made, never its transcript. The title's line says "in " and the Memory card says who the project is shared with. When a member leaves the project, their rooms leave it with them. Team projects. ## Caps - People: 12 in a private room, 100 in a public one; past that a public link is view-only. - Open rooms: 10 open rooms with people in them per account. A room of one does not count until someone is invited. - The purse: 100,000 credits at most in a room. - Threads: 50 a room. - A line: 24,000 characters, or up to the model's own window when it is larger. Up to 3 pictures and 6 files on a line. - Pictures: Up to 4 pictures a line on the Image door. - Compare: 2 to 4 models, each once. - Reactions: 24 on a line. ## Errors a room shows A refusal in a room is plain words, most with a code from the list. The room's own wordings: "This collab is closed.", "This collab is full, 12 people. You can still watch.", "You can have 10 open collabs with people in them. Close one first.", "A collab holds at most 100,000 credits.", "A collab holds 50 threads.", "Compare runs on the Ask and Code doors.", "Another model in this compare could not start, so none did.", "This thread runs no skill." and "Your memory stays out of a public room." Short of credits, a room of one says "Not enough credits for this run (about 52 needed). Top up your balance, or pick a cheaper model. (code W301)" and a room with people in it "The collab balance cannot cover this run (about 52 credits needed). Add credits to the collab. (code W301)"; a compare names the runs it could not start. None of them charge anything. | Code | What you see | What to do | | --- | --- | --- | | W301 | Not enough credits for this turn. | Top up, or pick a cheaper model. In a collab that pays from the room balance, add credits to the room: any member can. | | W825 | Not enough credits for this audio. | Top up, or pick a cheaper model. In a collab that pays from the room balance, add credits to the room: any member can. | | W101 | Please sign in to use the workspace. | Sign in again. | | W104 | Type a message first. | Type something. | | W106 | That message is too long for this model. | Pick a model with a larger window, such as Kimi K3 at 1 million tokens, or split it. | | W201 | That model is not in the catalogue right now. | Pick another model. | | W302 | This account is paused. | A crypto top-up pays what is owed first; a card payment waits until nothing is owed. Contact support if you think this is wrong. | | W514 | That model spent its whole output budget thinking. | Try again, shorten the prompt, or pick a model that thinks less. | | W606 | Too many attachments of one kind on a message. | Send them across several messages, or start a new chat. | | W607 | That attachment could not be found. | Attach the file again. | Source: https://heyaskr.ai/docs/workspace/collab --- # askr Skills What are the ten skills behind Ask, what does each one do, and how does the tier pick a model? Verified against the code on 2026-09-26. askr Skills are ten chats behind the Ask tile, each built for one job. The model writes plain words and small pieces of structured data; your device draws them as cards, calculators, steps and sources, checks them, and does the maths. Every skill runs on any model on askr, and each one opens on the model that suits it. ## The chooser Tap Ask and a popup opens: Ask anything, the plain chat, full width, then the ten skills two to a row, a sheet on a phone. Pick one and the chat opens on the skill's model, with its own message box hint, its own settings over the message box, and one line under the bar on what happens in it. A plain chat that asks about a loan, a symptom, a translation or a fix offers a chip to ask it in that skill, with the question ready to send: decided on your device from your words, no model call, easy to dismiss. On askr Pro, which a new chat opens on, a line that calls for a skill runs under it without asking: the answer arrives with the skill's screen under it in the same thread, and the pick line says which skill ran (askr Pro). ## The ten | Skill | What happens | Searches the web | Reads live | | --- | --- | --- | --- | | Look things up | askr searches the web for each question, answers from the pages it finds, and shows every source with its quote beside the chat. | Every turn | Nothing | | Write and edit | Paste text or say what to write. The draft sits beside the chat with the changes marked, and one tap makes it shorter, friendlier or more formal. | No | Nothing | | Learn and homework | Send a problem or a topic. askr guides you a step at a time, shows the working when you ask for it, and quizzes you on it after. | No | Nothing | | Fix it and how-to | Say what is broken and set your device. askr walks you through one checked step at a time and stops you before anything risky. | In Performance | Nothing | | Health and fitness | Ask about a result, a medicine or a goal. askr explains it, builds your appointment list and says when to get care. Information, not medical advice: in an emergency call 999 or 112. | In Performance | The matching MedlinePlus topic | | Translate | Type, paste or photograph text. Set the languages and the tone, tap a word for other choices, and check the meaning back. | No | Nothing | | Money | Ask about a loan, a budget or your bills. The calculators run on your device with the maths shown. Education, not financial advice. | In Performance | The Bank of England Bank Rate, UK inflation, exchange rates, and crypto prices when you ask | | Brainstorm | Say what the ideas are for and one limit. Each round arrives as cards to keep, drop, combine and build on. | No | Nothing | | Summarise | Paste text, a link or a document. You get the short version at the length you choose, with the quote behind each point. | Every turn | Nothing | | Shop and compare | Say what you are buying. askr asks what matters to you, searches current prices and compares the options side by side. Prices change: check the shop before you buy. | Every turn | Nothing | Look things up, Summarise and Shop and compare search on every turn where the model can; Fix it, Health and Money search in Performance, since current facts matter there. Health and Money carry a rule you always see: information, not medical advice; education, not financial advice. In an emergency Health shows the care card on your device before the reply and tells the model to agree with it. ## The screen Beside the chat on a desktop, a sheet on a phone, each skill draws what it made. Look things up shows every page it read as a card with the exact words that back each sentence. Write and edit keeps the draft beside the chat with the changes marked, a word count and reading age worked out on your device, and one-tap rewrites. Learn and homework reveals a worked solution one step at a time with a hint on each, checks your arithmetic on the device, and marks a quiz there too. Fix it shows a Fix card: what you need before you start, the difficulty, the steps easiest first, a did-this-fix-it check under the current step, and any risky step locked until you have read its warning. Health draws each result on a bar against the range printed on your report, builds your appointment list to tick off on the day, and shows a care card that says when to get help. Translate gives a two-part card with tappable words and other choices, a clarifying question when the meaning could go two ways, and photo translation for menus and signs. Money runs loan, mortgage, savings, debt and budget calculators on your device with the maths shown, and reads a statement photo into spending by category. Brainstorm deals ideas as cards to keep, drop, combine and build on, with a lens to filter them. Summarise gives an overview card with the quote behind each point. Shop and compare asks what matters first, then puts the options side by side with the better one marked on every row. The panel opens the moment you send, in the skill's real shape and empty: the headings, the rows, the controls, one line on what is coming, and your own words. Nothing is invented. Ask about a mortgage with the numbers in your question and the calculator opens with them already in, free, while the model is still thinking. On a phone a slim bar appears at send, pulses while askr works, and opens the sheet. ## Settings pills Each skill's own settings sit over the message box as pills, then quick replies: your device and its system in Fix it, the languages and the tone in Translate, the level and whether askr guides or explains in Learn, your country and currency in Money, what matters to you in Shop and compare, the length and format in Summarise. Menus float over the chat instead of pushing it down. Every setting goes with every turn, and the choices carry to your next chat in that skill, except Shop's criteria and Money's calculator, which belong to one chat. Every turn also carries the date in your time zone and your language, so a skill answers for where you are. ## Cheap, Balanced or Performance In a skill the mode chip in the message box becomes a tier: Cheap, a cheaper model and short answers; Balanced, strong models at an everyday price; Performance, the strongest models, thorough answers and more searching. Tapping a tier switches the model; picking a model by hand moves the tier to match, or to Custom when no tier lists it. Balanced is the default, and the choice is remembered for that skill's next chat. The answer style and the Performance searches follow the model actually sent, so the label is always what runs. | Skill | Cheap | Balanced | Performance | | --- | --- | --- | --- | | Look things up | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | claude-sonnet-5, gpt-6-sol | claude-opus-5.5, claude-fable-5.1, gpt-6-sol | | Write and edit | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | claude-sonnet-5, gpt-6-sol | claude-fable-5.1, claude-opus-5.5, gpt-6-sol | | Learn and homework | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | gpt-6-sol, claude-sonnet-5 | gpt-6-astra-pro, claude-fable-5.1, gpt-6-sol | | Fix it and how-to | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | claude-sonnet-5, gpt-6-sol | claude-fable-5.1, claude-opus-5.5, gpt-6-sol | | Health and fitness | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | claude-sonnet-5, gpt-6-sol | claude-fable-5.1, claude-opus-5.5, gpt-6-sol | | Translate | deepseek/deepseek-v4.1-flash, gemini-3.7-flash, gpt-5.4-nano | gemini-3.7-flash, claude-sonnet-5 | claude-fable-5.1, claude-opus-5.5, gpt-6-sol | | Money | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | gpt-6-sol, claude-sonnet-5 | gpt-6-astra-pro, claude-fable-5.1, gpt-6-sol | | Brainstorm | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | claude-sonnet-5, gpt-6-sol | claude-fable-5.1, claude-opus-5.5, gpt-6-sol | | Summarise | deepseek/deepseek-v4.1-flash, gemini-3.7-flash, gpt-5.4-nano | gemini-3.7-flash, claude-sonnet-5 | claude-fable-5.1, claude-opus-5.5, gpt-6-sol | | Shop and compare | gemini-3.7-flash, deepseek/deepseek-v4.1-flash, gpt-5.4-nano | claude-sonnet-5, gpt-6-sol | claude-opus-5.5, claude-fable-5.1, gpt-6-sol | The first model of a list the catalogue has is the one a chat opens on; the picker stays open to change it. A skill turn costs more than a plain turn on the same model: the skill's instructions, your settings, any live data and any search go in as input tokens, and the receipt under each reply says what the turn came to. There are no estimates inside a skill. What a run costs. ## Live data Health reads the matching MedlinePlus topic, and says "from MedlinePlus.gov" when it draws on it; it never promises a drug interaction check, since none is available to it. Money reads the Bank of England Bank Rate, UK inflation and exchange rates, and crypto prices when you mention one. Each is fetched before the answer, cached, and given two and a half seconds, so a slow source never holds a turn up. Fetched text reaches the model as data it is told to quote, not obey. ## Checked on your device - A citation in Look things up must name a page the search actually read for that answer. One that does not is repaired to the real page behind the same [n], without its quote, or dropped. - A quote in Summarise must be words from the text you pasted, or it is removed. - The maths in Money, the arithmetic in Learn and the diff in Write run on your device, never in the model, at no cost. - Health's emergency screen runs on your device and on the server, so the reply's care card agrees with it. ## Double-check In Look things up, Learn, Health, Money and Summarise, once there is an answer, Double-check asks a model from a different family to look for mistakes in it, in facts, maths, dates or safety, and say plainly what it would change and why. Only when you tap it, priced as one turn on that model. ## Save or share Whatever the skill made, the draft, the steps, the appointment list, the shortlist, the comparison, the sources, copies as text, saves as a file, or goes to your phone's share sheet. Fix it, Health, Translate, Money and Learn put a camera button first, for a photo of the label, the report, the menu, the statement or the question; Summarise and Health also take a PDF or a document. Attach files. ## A skill is a room A skill opens as a room whose thread carries the skill: the first send makes the room under the skill with your settings, and every line in that thread runs with the skill's instructions after the room's memory. The banner at the head of the thread names the skill; Change opens the chooser, and a pick starts a new thread in the same room under the new skill, while Ask anything starts a plain one. The settings live on the thread, so anyone in the room sees the same ones; the tier belongs to the thread and the model per line follows it. Compare, Re-run and Double-check work as in any room, and the room's memory notes a skill answer with the skill's name. The Image, Video and Audio doors are not offered inside a skill thread. askr Collab, askr Mind. ## How it works Each skill's instructions are held on our server and added to every turn of that chat, with the date, your language and your settings; no client can write them, and a request that tries is refused (code W105). The model answers in plain words plus fenced blocks of structured data with a kind, such as steps, draft, labs, calculator or citations. Your device cuts the blocks out of the reply and draws each with the skill's own component inside its own error boundary, so a bad block costs one card and never the chat; the rest is ordinary markdown, with maths typeset. Because the blocks are plain fenced JSON, every model on askr can run every skill. Source: https://heyaskr.ai/docs/workspace/skills --- # Your library Where do the things I make go, and who can see them? Verified against the code on 2026-09-15. Everything you make in the workspace that is not text is kept: images and video clips, each with its prompt, the model that made it, the credits it cost and when. Conversations are kept too, on your account; Chat says how, and the files you attach are on the Files page. Until askr has copied a finished clip, and for up to 24 hours regardless, it also exists at a temporary address on the model gateway. Your library is yours to see and nobody else's: nothing in it is published, browsed or used for anything but showing it back to you. Ask and it is deleted. Privacy. - Images shown: Your latest 100 - Clips shown: Your latest 60 - Who can see them: You, signed in. Files are served from askr and are not public URLs. - Download: Every item has a download that saves the original file. Pictures made through the API are not saved here; download them from the response. Images from chat. Source: https://heyaskr.ai/docs/workspace/library --- # Switching models How do I compare models on the same question, and what does the picker show? Verified against the code on 2026-09-15. The point of one balance is that trying another model costs nothing extra. In the workspace the model is a picker above the conversation; switch it and the next turn goes to the new model with the same history. ## What the picker shows - The name, the provider, and what one typical turn costs (1,000 tokens in, 500 out), so a cheap model and an expensive one are comparable at a glance. - Sorted cheapest first. The popular models are marked. - Only models that can run. Video models are in the video door, audio and embedding models are not offered, and a model whose provider is refusing to serve it is left out. ## A sensible default set | Want | Try | | --- | --- | | A strong all-rounder | gpt-5.6-sol | | Careful long writing and code | claude-sonnet-5 | | Fast and cheap for everyday questions | gemini-3.7-flash | | Cheapest capable reasoning | deepseek/deepseek-v4.1-flash | | A picture | google/gemini-3.1-flash-image | The catalog has the full list with live prices and context sizes. Source: https://heyaskr.ai/docs/workspace/models --- # What a run costs How is the price of a turn, a picture or a clip worked out? Verified against the code on 2026-09-15. One dollar is 1,000 credits. Each request is priced from what it actually cost upstream. There is no askr fee on usage today, so the credits you see charged are the model's own price. | Kind | Priced by | Shown | | --- | --- | --- | | Chat reply | Tokens in at the model's input rate, tokens out at its output rate | Before: the typical-turn price in the picker. After: the exact credits on the reply. | | Picture | The measured cost of a picture on that model | Before: the measured floor. After: the real figure. | | Video clip | The model's published price for that shape and length | Before, exactly. A clip has one price. | ## Why the picker price and the charge differ The picker prices a typical turn so models are comparable. Your actual turn has its own token count, and the model's reply length is not known until it finishes. The charge is always the real number, and it is what you see on the reply and in the wallet. ## Reserved, then settled When you send, the worst case is reserved from your balance so you can never start something you cannot afford. When the reply finishes, the reservation becomes the real charge. The wallet shows both your balance and what is available after reservations. Holds and settlement. ## Thinking models On a model that reasons before it answers, much of a short reply can be reasoning you never see. It is billed like any other output token. The picker's per-turn price does not include it, so expect these models to run above their listed turn price. Source: https://heyaskr.ai/docs/workspace/costs --- # Wallet, keys and activity What is in the wallet, and what does each number mean? Verified against the code on 2026-09-24. https://heyaskr.ai/wallet is the account page: your balance, what you have spent it on, how to add more, and your API keys. ## The numbers - Balance: Credits on the account - Available: Balance minus anything reserved by a request still running. This is what you can spend right now. - Usage value: What the balance is worth as model usage, in dollars. Deliberately not a cash value: credits cannot be withdrawn. ## Activity The last 50 entries on your ledger: deposits, every chat turn, every picture, every clip, with the model, the credits and the time. API calls appear here too, with the key that made them. There is no usage endpoint on the API; askr.credits_charged on each response is the per-call figure. ## Add credits Pick an asset and an amount, send the exact quote to the address shown. Add credits. ## Send credits You can send credits to any other @ on askr. In the workspace open Settings, then Balance & deposits (https://heyaskr.ai/chat/settings/balance), and press Send credits beside Top up. Type their @ and the dialog shows their picture and @ before you confirm. Enter a whole number of credits and, if you like, a note. The credits are on their balance the moment the send goes through, you get a receipt with a ref, they get an email if their account has one, and both of you see the send in the Sends card under the balance. - What can be sent: Your balance: credits you bought with crypto, credits askr gave you, and credits another @ sent you. Whole credits only. - What stays put: Promo credits from an airdrop, and the holder discount. Neither is part of the balance, so neither moves; promo keeps its own expiry and the discount follows your linked wallet. Credits bought by card are on the balance but never move either, and Sendable now leaves them out. Pay by card. - Who can receive: Any active account with an @. An email or wallet account that has not picked an @ yet cannot receive until it does. Referrals says how to pick one. - Minimum: 100 credits a send. - No limit: There is no cap on how much an account can send or receive, and none on how often. What you can send is what Sendable now shows. - Fresh deposits: Credits from a deposit made in the last 24 hours cannot be sent yet. Sendable now in the dialog is your balance minus those, minus anything a running request has reserved. - The code step: One send above 5,000 credits asks for a six-digit code, emailed to the account's address, good for ten minutes. An account with no email has no code step; its sends go through without one. - Final: Sends are final. There is no undo, and the recipient cannot decline. Sent to the wrong @, or a send you did not make? Email support with the ref on the receipt: askr can reverse a mistaken or fraudulent send on request, and the reversal shows on both sides. - Still credits: Sending changes whose balance the credits sit on and nothing else. On either side they buy model usage only and cannot be redeemed for money or crypto. Refunds and withdrawals. A send the dialog refuses says why in plain words, with a code from W901 to W913, or W302. They are listed on Error codes. ## API keys Make, cap and revoke keys. A key is shown once. Authentication and keys. ## Signing out Sign out ends this browser's session. Sign out everywhere ends every session on the account, which is the thing to press if a device is lost. Source: https://heyaskr.ai/docs/workspace/wallet --- # The API What the API is, where it lives, and what is on it. Verified against the code on 2026-09-24. The askr API is the OpenAI chat completions API with a different base URL and key. Anything that speaks that shape works: Cursor, Continue, Aider, the OpenAI SDKs, LangChain, or a curl command. The surface is deliberately small. - Base URL: https://heyaskr.ai/v1 - Auth: Authorization: Bearer askr_live_... - Format: JSON in, JSON out; server-sent events when stream: true - Money: Credits. One dollar is 1,000 credits. Every response says what it cost. ## The endpoints | Method | Path | What it does | | --- | --- | --- | | GET | /v1 | Connection check. Your balance and the key in use. Detail | | GET | /v1/models | Every chat model you can call, in the shape the OpenAI SDKs expect, with what each takes beside words. Detail | | POST | /v1/chat/completions | The one that does the work. Streaming and non-streaming; pictures and files in the message. Detail | | POST, GET, DELETE | /v1/files | Upload a file once, then name it by file_id in any message. List, fetch, delete. Detail | ## What is not on it Video runs in the workspace on its own endpoint and is not on the API yet. Audio and embedding models are refused. There are no images, assistants, fine-tuning or responses endpoints. Pictures come from chat models that answer with an image, in message.images; pictures, PDFs, sound and clips go in as message parts or by file_id. The full list, with where each item sits on the roadmap, is on What is not on the API yet. ## Where it differs from OpenAI The request and response shapes match. There are five deliberate differences, and Differences from OpenAI lists them. The one you will use: askr.credits_charged on every response. ## Browsers The API sends no CORS headers, so a web page on another origin cannot call it directly, and it should not: a key in front-end code is a key anyone can spend. Call it from your server, a CLI, a desktop tool or a script. - Get a key: Made in the wallet, shown once, with an optional daily cap. - Your first request: curl, Python and Node, ready to paste. - Errors: Every status we return, and whether it charged you. Source: https://heyaskr.ai/docs/api --- # Authentication and keys How do I get a key, what can it do, and how do I keep it safe? Verified against the code on 2026-09-15. Every API call carries a key as a bearer token. Keys are made in the wallet and belong to your account, so anything they spend comes off your balance. ```http Authorization: Bearer askr_live_... ``` ## Making a key 1. Open the wallet. https://heyaskr.ai/wallet#api, signed in. 2. Name it, optionally cap it. One key per app or machine is a good habit. A daily cap, in credits, bounds what that key can spend in any rolling 24 hours. 3. Copy it now. The key is shown once. Only a hash is kept, so it cannot be recovered later, only revoked and replaced. ## Facts about keys - Format: askr_live_ followed by 32 URL-safe characters - Storage: SHA-256 hash only; the plain key exists nowhere after you copy it - Active keys: Up to 20 per account - Creation rate: 10 per hour - Daily cap: Optional, in credits, checked before each request against the worst case that request could cost. Daily caps - Scopes: None. Every key can do everything the API does. - Expiry: None. A key works until you revoke it. - Revoking: Immediate. The next request with that key fails with 401. Its past usage stays in your activity. ## When a key is wrong ```json { "error": { "message": "Missing or invalid API key.", "type": "authentication_error", "param": null, "code": null } } ``` That is the answer for a missing header, a revoked key and a truncated paste alike. Make a new key rather than debugging an old one. ## Keep it like a card number - Environment variable or a secrets store, never a committed file and never front-end code. - Set a daily cap on any key a script uses. A retry loop can spend a balance in minutes in a way nobody typing into a chat box ever will. - One key per place. Revoking one then costs you one place, not all of them. ## Test a key ```bash curl https://heyaskr.ai/v1 -H "Authorization: Bearer $ASKR_KEY" ``` That prints your balance and the key in use. Connection check. Source: https://heyaskr.ai/docs/api/authentication --- # Chat completions Every field we read, every limit we apply, and what comes back. Verified against the code on 2026-09-24. ```http POST https://heyaskr.ai/v1/chat/completions ``` Create a model response for a conversation. The request is the OpenAI shape. This page is precise about which fields we read, because the rest are dropped without a warning. ## Request fields we read | Field | Type | What it does | | --- | --- | --- | | model | string | A chat model id from the catalog. Exact match; there are no suffixes or aliases. | | messages | array | The conversation. Each item has role (system, user or assistant) and content: a string, or an array of parts for pictures and files. Pictures and files. | | max_tokens | integer | Output ceiling, thinking included. Default 4,096, maximum 8,192. Larger values are clamped, not rejected. | | stream | boolean | true for server-sent events. Streaming. | | web_search | boolean | askr's own field. true lets the model search the web and read pages during the call; the sources come back as annotations. Search the web. | > Everything else is dropped silently: temperature, top_p, tools, tool_choice, response_format, stop, n, seed, plugins. A request that sends them succeeds; they just have no effect. Tool calling and structured output are on the roadmap. Web search is not a tool you define: it is the web_search field above. ## Messages - Roles other than system, user and assistant return 400 Unsupported message role. - content is a string, or an array of parts: text, image_url, file, input_audio and video_url, each inline or by file_id. Any other part type returns 400 Unsupported content part. A message that is only parts is sent as it is. Pictures and files. - Only the last 40 messages are sent to the model. Trim on your side if you want control over what is kept. - A conversation may fill the model's own context window, less room for the reply: about 1 million tokens on the largest models, such as Kimi K3, at most 3.2 million characters and the last 400 messages. A model with a small window keeps 120,000 characters and 40 messages. Over the limit returns 400 The conversation is too long. - No usable messages returns 400 No usable messages were provided. ## Example ```bash curl https://heyaskr.ai/v1/chat/completions \ -H "Authorization: Bearer $ASKR_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "claude-sonnet-5", "max_tokens": 300, "messages": [ {"role": "system", "content": "Answer in one sentence."}, {"role": "user", "content": "Why is the sky blue?"} ] }' ``` ```python from openai import OpenAI client = OpenAI(base_url="https://heyaskr.ai/v1", api_key="askr_live_...") resp = client.chat.completions.create( model="claude-sonnet-5", max_tokens=300, messages=[ {"role": "system", "content": "Answer in one sentence."}, {"role": "user", "content": "Why is the sky blue?"}, ], ) print(resp.choices[0].message.content) print(resp.model_extra["askr"]["credits_charged"]) ``` ```javascript import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://heyaskr.ai/v1", apiKey: "askr_live_..." }); const resp = await client.chat.completions.create({ model: "claude-sonnet-5", max_tokens: 300, messages: [ { role: "system", content: "Answer in one sentence." }, { role: "user", content: "Why is the sky blue?" }, ], }); console.log(resp.choices[0].message.content, resp.askr.credits_charged); ``` ## Response ```json { "id": "chatcmpl-9c2b...", "object": "chat.completion", "created": 1789516800, "model": "claude-sonnet-5", "choices": [ { "index": 0, "message": { "role": "assistant", "content": "Sunlight scatters off air molecules, and blue scatters most." }, "finish_reason": "stop" } ], "usage": { "prompt_tokens": 21, "completion_tokens": 14, "total_tokens": 35, "completion_tokens_details": { "reasoning_tokens": 0 } }, "askr": { "credits_charged": 0.192 } } ``` | Field | Meaning | | --- | --- | | choices[0].message.content | The reply. Always one choice. | | choices[0].message.images | Present only when the model answered with a picture. Images from chat. | | choices[0].message.annotations | Present only on a web_search: true call: one url_citation per source. Search the web. | | usage.completion_tokens_details.reasoning_tokens | On a thinking model, output tokens you never see. They are billed like any other output token. | | askr.credits_charged | What the call cost, in credits, to four decimal places. Additive: clients that do not know it ignore it. | ## How the cost is worked out Before the call, credits are reserved for the worst case: your estimated input tokens at the model's input rate, plus max_tokens at its output rate. When the model finishes, the reservation is settled to the real cost, using the gateway's own figure for the call. Nothing else is added: there is no platform fee in the first version. Holds and settlement. ## Pictures and files in a message A user message's content can be an array of parts, in OpenAI's shape. Five kinds are read. A part carries its bytes inline, or names an upload by file_id from POST /v1/files; an upload is the way to send anything big, or anything more than once. | Part | Inline | By upload | Inline limit | | --- | --- | --- | --- | | text | text | | | | image_url | image_url.url: a PNG, JPEG, WebP or GIF data URL | image_url.file_id | 8 MB decoded; 6 a message, 20 a request | | file | file.file_data: a PDF as a data URL, or an https URL the provider fetches. file.filename is optional. | file.file_id | 32 MB decoded | | input_audio | input_audio.data as base64, with input_audio.format wav or mp3 | input_audio.file_id | 25 MB decoded | | video_url | video_url.url: an MP4, MOV or WebM data URL, or an https URL | video_url.file_id | 25 MB decoded | The request body on this endpoint is 12 MB, so an inline part over about 9 MB does not fit: upload it. An inline part goes to the model as it is, so it works only on a model whose inputs on GET /v1/models include that kind. A part by file_id is routed per model: the file itself where the model takes the kind, its words where it does not, so a PDF by file_id works on every model. Files has the rules and the costs. ```json { "role": "user", "content": [ { "type": "text", "text": "What do these say, and does the clip match?" }, { "type": "file", "file": { "file_id": "3f2a9c1e-7b4d-4e0a-9c1f-2d8e6a5b4c3d" } }, { "type": "file", "file": { "filename": "notes.pdf", "file_data": "data:application/pdf;base64,JVBERi0xLjcK..." } }, { "type": "input_audio", "input_audio": { "data": "UklGRiQAAABXQVZF...", "format": "wav" } }, { "type": "video_url", "video_url": { "url": "data:video/mp4;base64,AAAAIGZ0eXBpc29t..." } } ] } ``` ```bash curl https://heyaskr.ai/v1/chat/completions \ -H "Authorization: Bearer $ASKR_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "claude-sonnet-5", "messages": [{"role": "user", "content": [ {"type": "text", "text": "Summarise this in five bullets."}, {"type": "file", "file": {"file_id": "3f2a9c1e-7b4d-4e0a-9c1f-2d8e6a5b4c3d"}} ]}] }' ``` ```python from openai import OpenAI client = OpenAI(base_url="https://heyaskr.ai/v1", api_key="askr_live_...") f = client.files.create(file=open("report.pdf", "rb"), purpose="user_data") resp = client.chat.completions.create( model="claude-sonnet-5", messages=[{"role": "user", "content": [ {"type": "text", "text": "Summarise this in five bullets."}, {"type": "file", "file": {"file_id": f.id}}, ]}], ) print(resp.choices[0].message.content) ``` Cost. A file is input tokens at the model's input rate. The hold reserves 1,600 tokens a picture, 1,600 a PDF page, 32 a second of sound, 100 a second of video, and the text's own count where a file goes as words; it settles to the gateway's real figure. The three newest user messages that name uploads carry them in full; older ones get a one-line note in the file's place, so a PDF at the top of a long conversation is not paid for on every call. Refused before anything runs. Nothing is reserved for these. Errors. | Status | Message | Why | | --- | --- | --- | | 400 | The model 'x' does not accept image input. | An inline picture on a model whose inputs lack image. | | 400 | The model 'x' does not accept file input. Upload the file with POST /v1/files and send its file_id instead; askr then sends the model its text where it can. | An inline file, input_audio or video_url part on a model that takes no such input. By file_id, a PDF or sound goes as words instead. | | 400 | A file part needs file.file_id (an upload), or file.file_data as a PDF data URL or an https URL. | The part is malformed. input_audio and video_url parts say the same in their own words. | | 404 | No file with id 'x'. | The file_id names nothing on your account: deleted, swept after 7 days unused, or someone else's. | | 413 | At most 6 pdf files per request. | More than the per-request count for that kind, across every message sent. Limits. | | 413 | This request, with its files, is about 180,000 tokens; the model 'x' takes 128,000. | Over the model's window. code is context_length_exceeded. Pick a bigger window, or send less. | | 400 | That model cannot watch video. Pick one that does, or remove the video. | A workspace code in code: W609 here; also W203, W605, W606, W607, W608 and W611. Error codes. | ## Limits - Requests: 120 per minute per IP address - Body size: 12 MB on this endpoint, for inline pictures and files; uploads go to /v1/files at up to 100 MB - Output: max_tokens up to 8,192; default 4,096 - Deadline: 120 seconds upstream. Past it you get 504 and the reservation is charged, because the model did run. Errors ## Search the web Send "web_search": true and the model can search the web and read pages during the call, exactly as the globe in the workspace does. The model is given two tools of ours, web_search and fetch_page, decides for itself whether to use them, and answers with numbered [n] marks. Per call: at most 3 searches, 2 pages read and 4 model calls, the last of them made without tools. It works on a model whose entry on GET /v1/models says supportsTools: true, streamed or not. Your own tools are still dropped. ```bash curl https://heyaskr.ai/v1/chat/completions \ -H "Authorization: Bearer $ASKR_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-5.6-sol", "web_search": true, "messages": [{"role": "user", "content": "When did askr add web search?"}] }' ``` ```python from openai import OpenAI client = OpenAI(base_url="https://heyaskr.ai/v1", api_key="askr_live_...") resp = client.chat.completions.create( model="gpt-5.6-sol", messages=[{"role": "user", "content": "When did askr add web search?"}], extra_body={"web_search": True}, ) print(resp.choices[0].message.content) for a in resp.choices[0].message.annotations or []: print(a.url_citation.url) print(resp.usage.model_extra["searches"], resp.model_extra["askr"]["credits_charged"]) ``` ```javascript import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://heyaskr.ai/v1", apiKey: "askr_live_..." }); const resp = await client.chat.completions.create({ model: "gpt-5.6-sol", messages: [{ role: "user", content: "When did askr add web search?" }], web_search: true, // askr's field; cast the params in TypeScript }); console.log(resp.choices[0].message.content); for (const a of resp.choices[0].message.annotations ?? []) console.log(a.url_citation.url); console.log(resp.usage.searches, resp.askr.credits_charged); ``` ```json { "id": "chatcmpl-9c2b...", "object": "chat.completion", "created": 1789516800, "model": "gpt-5.6-sol", "choices": [ { "index": 0, "message": { "role": "assistant", "content": "Search the web went live on 23 September [1].", "annotations": [ { "type": "url_citation", "url_citation": { "url": "https://heyaskr.ai/changelog", "title": "Changelog", "start_index": 41, "end_index": 44 } } ] }, "finish_reason": "stop" } ], "usage": { "prompt_tokens": 3412, "completion_tokens": 58, "total_tokens": 3470, "searches": 1, "fetches": 0, "search_credits": 5 }, "askr": { "credits_charged": 5.31 } } ``` | Field | Meaning | | --- | --- | | message.annotations | One per source the call read, in the order the answer numbers them: [1] is annotations[0]. The shape is OpenAI's url_citation, so a client that reads OpenAI's search results reads ours. | | url_citation.url, url_citation.title | The page, and its title. | | url_citation.start_index, url_citation.end_index | The span of the first [n] mark in content, in UTF-16 code units as OpenAI counts them. A source the model listed but never cited inline has an empty span at the end of the text. | | usage.searches, usage.fetches | How many searches ran and how many pages were read. Zero and zero when the model chose not to search. | | usage.search_credits | The searches at 5 credits each. | | askr.credits_charged | The whole charge: the model's tokens across every call plus search_credits. | Streamed, the chunks are the ones on Streaming, with three differences: content from every model call arrives in order with a blank line between calls; a : askr comment line appears while a search or a page read runs, so an idle proxy does not close the stream; and the last data frame carries delta.annotations, finish_reason, usage and askr together. ```text data: {"id":"chatcmpl-9c2b...","object":"chat.completion.chunk","created":1789516800,"model":"gpt-5.6-sol","choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null}]} : askr data: {"id":"chatcmpl-9c2b...","object":"chat.completion.chunk","created":1789516800,"model":"gpt-5.6-sol","choices":[{"index":0,"delta":{"content":"Search the web went live"},"finish_reason":null}]} ... data: {"id":"chatcmpl-9c2b...","object":"chat.completion.chunk","created":1789516800,"model":"gpt-5.6-sol","choices":[{"index":0,"delta":{"content":"","annotations":[{"type":"url_citation","url_citation":{"url":"https://heyaskr.ai/changelog","title":"Changelog","start_index":41,"end_index":44}}]},"finish_reason":"stop"}],"usage":{"prompt_tokens":3412,"completion_tokens":58,"total_tokens":3470,"searches":1,"fetches":0,"search_credits":5},"askr":{"credits_charged":5.31}} data: [DONE] ``` Cost. Each search is 5 credits. The text the model reads is billed as input tokens at the model's rate, like the rest of the conversation, and every extra model call sends the conversation again. The hold is raised by 15 credits plus the most text the searches, the pages and the extra calls could add at the model's input rate; a 402 names that figure. It settles once, to the searches that ran plus the real token usage across every call. A call cut short is settled on what streamed, at most the hold. A call that ends with no content releases the hold and charges nothing, searches included. Each account may start 6 searching calls a minute, workspace and API together. Refused before anything runs. Nothing is reserved for these. All three carry "param": "web_search" and the workspace's code in code, so one code means one thing on both surfaces. Errors. | Status | Type | Message | Why | | --- | --- | --- | --- | | 400 | invalid_request_error | web_search is not available on this server. (code W206) | This server has no search. Drop web_search. | | 400 | invalid_request_error | This model cannot use web_search; see supportsTools on GET /v1/models. (code W207) | The model takes no tools, or answers with pictures. Pick one with supportsTools: true. | | 429 | rate_limit_error | Search the web is limited to a few turns a minute. Wait a moment and send again. (code W109) | More than 6 searching calls in a minute on the account. Wait and retry. | Source: https://heyaskr.ai/docs/api/chat-completions --- # Streaming How do I get tokens as they are generated, and what does stopping early cost? Verified against the code on 2026-09-23. Send "stream": true and the response is a standard server-sent event stream, the same frames your client already parses. The proxy in front of the API does not buffer, so the first token arrives when the model produces it. ```bash curl -N https://heyaskr.ai/v1/chat/completions \ -H "Authorization: Bearer $ASKR_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"gpt-5.6-sol","stream":true,"messages":[{"role":"user","content":"Write a haiku about the sea."}]}' ``` ```python from openai import OpenAI client = OpenAI(base_url="https://heyaskr.ai/v1", api_key="askr_live_...") stream = client.chat.completions.create( model="gpt-5.6-sol", messages=[{"role": "user", "content": "Write a haiku about the sea."}], stream=True, ) for chunk in stream: delta = chunk.choices[0].delta.content if chunk.choices else None if delta: print(delta, end="", flush=True) if chunk.usage: print("\ncost:", chunk.model_extra["askr"]["credits_charged"]) ``` ```javascript import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://heyaskr.ai/v1", apiKey: "askr_live_..." }); const stream = await client.chat.completions.create({ model: "gpt-5.6-sol", messages: [{ role: "user", content: "Write a haiku about the sea." }], stream: true, }); for await (const chunk of stream) { process.stdout.write(chunk.choices[0]?.delta?.content ?? ""); if (chunk.usage) console.log("\ncost:", chunk.askr.credits_charged); } ``` ## What the wire looks like ```text : askr data: {"id":"chatcmpl-9c2b...","object":"chat.completion.chunk","created":1789516800,"model":"gpt-5.6-sol","choices":[{"index":0,"delta":{"role":"assistant","content":"Salt"},"finish_reason":null}]} data: {"id":"chatcmpl-9c2b...","object":"chat.completion.chunk","created":1789516800,"model":"gpt-5.6-sol","choices":[{"index":0,"delta":{"content":" wind"},"finish_reason":null}]} ... data: {"id":"chatcmpl-9c2b...","object":"chat.completion.chunk","created":1789516800,"model":"gpt-5.6-sol","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]} data: {"id":"chatcmpl-9c2b...","object":"chat.completion.chunk","created":1789516800,"model":"gpt-5.6-sol","choices":[],"usage":{"prompt_tokens":14,"completion_tokens":19,"total_tokens":33},"askr":{"credits_charged":0.23}} data: [DONE] ``` - Lines starting with : are keepalive comments. Ignore them. - Frames are re-encoded to the OpenAI chunk shape. Provider-specific fields and reasoning deltas are not passed through. - The last data frame before [DONE] carries usage and askr.credits_charged. - Pictures from image-capable chat models arrive as delta.images. Images from chat. - With web_search: true, a : askr comment appears while a search runs and the last data frame also carries delta.annotations and finish_reason. Search the web. ## Stopping early Close the connection and the upstream call is cancelled immediately. You are charged for what was generated up to that point and nothing after it. The final usage frame will not arrive, so read the cost from your activity or the next GET /v1. ## Headers we send ```http Content-Type: text/event-stream; charset=utf-8 Cache-Control: no-cache, no-transform Connection: keep-alive X-Accel-Buffering: no ``` Source: https://heyaskr.ai/docs/api/streaming --- # Images from chat How do I generate a picture, and where does it come back? Verified against the code on 2026-09-15. There is no images endpoint. Some chat models answer with a picture, and on askr those pictures come back inside the chat completion. Ask in words, get an image URL. The models that do this are the chat models whose output includes image in the catalog: today the three Gemini image models (the ones the internet calls Nano Banana). The GPT image models are listed upstream but on the refused list, and catalog entries of type image that are not chat models are refused with 404. ```bash curl https://heyaskr.ai/v1/chat/completions \ -H "Authorization: Bearer $ASKR_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "google/gemini-3.1-flash-image", "messages": [{"role": "user", "content": "A red fox in a snowy forest, cinematic light."}] }' ``` ```python import requests r = requests.post( "https://heyaskr.ai/v1/chat/completions", headers={"Authorization": "Bearer askr_live_..."}, json={ "model": "google/gemini-3.1-flash-image", "messages": [{"role": "user", "content": "A red fox in a snowy forest, cinematic light."}], }, ).json() for image in r["choices"][0]["message"].get("images", []): print(image["image_url"]["url"]) print("cost", r["askr"]["credits_charged"]) ``` ```json { "choices": [{ "index": 0, "message": { "role": "assistant", "content": "", "images": [{ "type": "image_url", "image_url": { "url": "https://..." } }] }, "finish_reason": "stop" }], "usage": { "prompt_tokens": 18, "completion_tokens": 1290, "total_tokens": 1308 }, "askr": { "credits_charged": 83 } } ``` ## What a picture costs These models publish per-token rates that do not describe what a picture costs, because most of the output tokens are the picture. So the reservation uses a measured floor per model (41 to 163 credits on the Gemini image models, 200 for any we have not measured) and the charge is settled to the gateway's real figure. The cost is in askr.credits_charged as always. What you are charged when. ## Streaming Works. The picture arrives as delta.images on a chunk. If you want the simple path, do not stream for images: one response, one URL. ## Keeping the picture Download it. The URL is a temporary address on the model gateway and lives for up to 24 hours. Pictures made in the workspace are kept in your library; pictures made through the API are not. Source: https://heyaskr.ai/docs/api/images --- # List models What does the models endpoint return, and why is it shorter than the catalog? Verified against the code on 2026-09-24. ```http GET https://heyaskr.ai/v1/models ``` Returns every model you can call on the API, in the shape the OpenAI SDKs expect. A key is required. ```bash curl https://heyaskr.ai/v1/models -H "Authorization: Bearer $ASKR_KEY" ``` ```json { "object": "list", "data": [ { "id": "gpt-5.6-sol", "object": "model", "created": 0, "owned_by": "OpenAI", "supportsTools": true, "inputs": ["text", "image", "file"] }, { "id": "claude-sonnet-5", "object": "model", "created": 0, "owned_by": "Anthropic", "supportsTools": true, "inputs": ["text", "image", "file"] } ] } ``` ## What is in the list - Chat models only. Video, audio and embedding models are not callable here, so they are not listed. - Only models with a published rate. A model we cannot bill is not a model we can serve. - Not the ids whose provider is currently refusing to serve them. Listed but refused. - supportsTools on each entry says whether web_search: true works on it. Additive: SDKs that do not know the field ignore it. Search the web. - inputs lists what the model takes beside words: text, then any of image, file (a PDF), audio and video, as the catalogue says. An inline part in a message must be in this list. A file by file_id is routed by it: the file where the kind is listed, its words where it is not. Additive. Files. ## What is not in it Prices and context length. For those use the catalog page, which reads the same feed with everything attached, or the public JSON at /api/models, which needs no key. > There are no query filters on this endpoint. ?type=image and friends are ignored. Source: https://heyaskr.ai/docs/api/models --- # Files How do I upload a file once and send it to any model? Verified against the code on 2026-09-24. ```http POST https://heyaskr.ai/v1/files GET https://heyaskr.ai/v1/files GET https://heyaskr.ai/v1/files/{id} GET https://heyaskr.ai/v1/files/{id}/content DELETE https://heyaskr.ai/v1/files/{id} ``` Upload a file once, then name it by file_id in any message. The shape is OpenAI's files API, so client.files.create in the SDKs works as it is. Uploading is free and keeping a file costs nothing; a file costs input tokens on each call that carries it, at the model's rate. The same store sits behind the paperclip in the workspace, with the same limits and the same rules. ## Upload, then ask ```bash curl https://heyaskr.ai/v1/files \ -H "Authorization: Bearer $ASKR_KEY" \ -F "file=@report.pdf" \ -F "purpose=user_data" curl https://heyaskr.ai/v1/chat/completions \ -H "Authorization: Bearer $ASKR_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "claude-sonnet-5", "messages": [{"role": "user", "content": [ {"type": "text", "text": "Summarise this in five bullets."}, {"type": "file", "file": {"file_id": "3f2a9c1e-7b4d-4e0a-9c1f-2d8e6a5b4c3d"}} ]}] }' ``` ```python from openai import OpenAI client = OpenAI(base_url="https://heyaskr.ai/v1", api_key="askr_live_...") f = client.files.create(file=open("report.pdf", "rb"), purpose="user_data") print(f.id, f.model_extra["askr"]["meta"]) # {'pages': 12, 'chars': 30412} resp = client.chat.completions.create( model="claude-sonnet-5", messages=[{"role": "user", "content": [ {"type": "text", "text": "Summarise this in five bullets."}, {"type": "file", "file": {"file_id": f.id}}, ]}], ) print(resp.choices[0].message.content) print(resp.model_extra["askr"]["credits_charged"]) ``` The part to use for an upload follows its kind: file for a PDF, a document or a text file, image_url for a picture, input_audio for sound, video_url for a clip, each with file_id in place of the inline data. Pictures and files in a message has every shape. ## The five routes | Method | Path | What it does | | --- | --- | --- | | POST | /v1/files | Multipart with a file field. purpose is accepted and ignored: every upload here is for a model to read. An optional name field sets the filename kept. Answers with the file object. 60 uploads a minute. | | GET | /v1/files | Your newest 100 files, as { "object": "list", "data": [...] }. | | GET | /v1/files/{id} | One file object. | | GET | /v1/files/{id}/content | The bytes, with the stored content type and the filename. | | DELETE | /v1/files/{id} | { "id": ..., "object": "file", "deleted": true }. A call that names the id afterwards gets 404. | ## The file object ```json { "id": "3f2a9c1e-7b4d-4e0a-9c1f-2d8e6a5b4c3d", "object": "file", "bytes": 482113, "created_at": 1789516800, "filename": "report.pdf", "purpose": "user_data", "askr": { "kind": "pdf", "mime": "application/pdf", "meta": { "pages": 12, "chars": 30412 } } } ``` | Field | Meaning | | --- | --- | | askr.kind | What the bytes were found to be: image, pdf, text, document, audio or video. The name and the declared type are not trusted. | | askr.mime | The stored type, from the bytes. | | askr.meta | What was measured: width and height for a picture; pages for a PDF; seconds for audio and video; sheets or slides for a document; chars for anything read as words, with truncated: true when it was cut at 400,000 characters. | ## What is accepted | Kind | Formats | Per file | Per message | Per request | | --- | --- | --- | --- | --- | | Picture | PNG, JPEG, WebP, GIF | 8 MB, 8,192 pixels a side | 6 | 20 | | PDF | PDF | 32 MB | 3 | 6 | | Text and code | Plain text, Markdown, CSV, JSON, HTML, YAML and source code in most languages | 5 MB | 5 | 10 | | Document | Word (.docx), PowerPoint (.pptx), Excel (.xlsx) | 32 MB | 3 | 6 | | Audio | MP3, WAV, M4A, OGG, FLAC, WebM | 25 MB, 20 minutes | 2 | 4 | | Video | MP4, MOV, WebM | 100 MB, 10 minutes | 1 | 2 | Per request counts every file_id on every message in messages. A kind not in the table is refused with 415, and an empty or damaged file with 400, at upload. Pictures are not resized on the API: send them at the size you want the model to see, under 8 MB. ## How a file reaches the model Every model reads words. Only some read files. A file by file_id goes to the model as the file where the model's inputs on GET /v1/models include its kind, and as its words where they do not: - Pictures go to a model with image in its inputs and are refused on any other (W203). 212 chat models see pictures, on the catalogue of 24 September 2026. - PDFs of up to 100 pages go as the file to a model with file in its inputs, 113 of them: inline under 8 MB, by a link the provider fetches above it. On every other model, and over 100 pages on every model, a PDF goes as its text. A scanned PDF with no text layer works only on a model that reads files (W611 otherwise). - Text, code and documents always go as words, between lines that name the file. Word is read as its paragraphs, PowerPoint one block per slide, Excel one sheet at a time as CSV. Up to 400,000 characters a file, about 100,000 tokens; a longer file is cut there and the model is told. - Audio goes as sound to a model with audio in its inputs, 32 of them, when it is an MP3 or a WAV under 8 MB. Anything else, and any audio for a model that cannot hear, goes as a transcript, made once by a speech-to-text model and kept beside the file. If none can be made, W608. - Video goes as the clip to a model with video in its inputs, 66 of them: inline under 8 MB, by a link above it. There is no words version of a video; on any other model it is refused (W609). ## What it costs A file on a call is input tokens at the model's input rate. The hold reserves the top of what providers bill per unit and settles to the gateway's real figure, in askr.credits_charged as always. Holds and settlement. | Sent as | Reserved | | --- | --- | | A picture | 1,600 tokens | | A PDF, as the file | 1,600 tokens a page | | Audio, as sound | 32 tokens a second | | Video | 100 tokens a second | | Words: a text file, a document, a PDF's text, a transcript | The text's own token count | Only the three newest user messages that name uploads carry them in full. On an older message the file is replaced by a one-line note saying it was attached there, so a client that resends the whole conversation, as OpenAI-shaped clients do, does not pay for a PDF on every call. Files count toward the model's window with everything else; a request that does not fit returns 413 with code context_length_exceeded before anything runs. ## How long a file is kept - Used in a call: Kept on your account until you delete it. The workspace's Files page shows the same list. - Never used: Swept 7 days after upload. - The words read from it: Extracted text and any transcript are kept beside the file, so they are made once. - Where it goes: Where your words go: to the model's provider, through askr's gateway, for that call. A file over 8 MB sent as the file is fetched by the provider from a link that expires after an hour. ## Errors None of these charge anything. Refusals that share a message with the workspace carry its code in code, so Error codes reads for both. | Status | Where | Message | Why | | --- | --- | --- | --- | | 400 | Upload | Send multipart/form-data with a file field. | The body was not multipart. | | 415 | Upload | That file type is not accepted. Pictures (PNG, JPEG, WebP, GIF), PDFs, text and code files, Word, PowerPoint and Excel documents, audio (MP3, WAV, M4A, OGG) and video (MP4, MOV, WebM) are. | code is W604. Judged from the bytes. | | 413 | Upload | That file is too large. The limit is 32 MB for a PDF. | code is W605. The message names the limit for that kind. | | 400 | Upload | That file could not be read. It may be empty or damaged. | code is W610. | | 404 | Files, chat | No file with id 'x'. | Deleted, swept, or on another account. | | 413 | Chat | At most 6 pdf files per request. | More than the per-request count for that kind. | | 400 | Chat | The model 'x' does not accept file input. Upload the file with POST /v1/files and send its file_id instead; askr then sends the model its text where it can. | An inline part on a model that takes no such input. Also image, audio and video. | | 400 | Chat | That model cannot watch video. Pick one that does, or remove the video. | A per-model refusal of an upload: W609 here; also W203, W605, W606, W607, W608, W611. | | 413 | Chat | This request, with its files, is about 180,000 tokens; the model 'x' takes 128,000. | code is context_length_exceeded. Pick a bigger window, or send less. | Source: https://heyaskr.ai/docs/api/files --- # Connection check How do I confirm a key works and see my balance from a script? Verified against the code on 2026-09-15. ```http GET https://heyaskr.ai/v1 ``` One call that says whether the API is up, whether your key is valid, and how many credits are on the account. It answers without a key too, minus the account. ```bash curl https://heyaskr.ai/v1 -H "Authorization: Bearer $ASKR_KEY" ``` ```bash # curl gets a terminal banner. Any other User-Agent gets JSON: curl https://heyaskr.ai/v1 -A "my-script" -H "Authorization: Bearer $ASKR_KEY" ``` ```json { "object": "askr.api", "base_url": "https://heyaskr.ai/v1", "models": 340, "streaming": true, "authenticated": true, "account": { "email": "you@example.com", "credits": 12480.5 }, "endpoints": ["/v1/models", "/v1/chat/completions"], "docs": "https://heyaskr.ai/docs/api" } ``` | Field | Meaning | | --- | --- | | models | How many chat models currently have a rate and can be called | | authenticated | Whether the key you sent was accepted | | account.credits | Balance in credits. One dollar is 1,000. | ## Why curl gets a banner When the User-Agent is curl, wget, httpie, PowerShell or fetch, the response is plain text with colour: the email, the key's name and prefix, the balance and what it buys. That is for a person at a terminal. Set any other User-Agent and you get the JSON above. ## Balance after a call You rarely need this endpoint for that: every completion returns askr.credits_charged, so a script can keep its own running total and check in here only occasionally. Source: https://heyaskr.ai/docs/api/connection-check --- # Errors, and what each one charged What does each status mean, and did it cost me anything? Verified against the code on 2026-09-24. Errors use standard HTTP status codes and the OpenAI error envelope. The column that matters most is the last one: some failures happen after the model has already run, and those are charged. ```json { "error": { "message": "Not enough credits. This request needs about 12. Add credits at https://heyaskr.ai/wallet", "type": "insufficient_quota", "param": null, "code": null } } ``` ## Before the model runs None of these charge anything. The reservation, if one was made, is released. | Status | Type | Message | What to do | | --- | --- | --- | --- | | 400 | invalid_request_error | messages is required and must be a non-empty array. | Send a messages array. | | 400 | invalid_request_error | Unsupported message role: X | Use system, user or assistant. | | 400 | invalid_request_error | No usable messages were provided. | Send at least one message with string content. | | 400 | invalid_request_error | The conversation is too long. | Trim below the model's own context window (120,000 characters on a model with a small window). | | 400 | invalid_request_error | That model has no published rate and cannot be billed. | Pick a model from the catalog. | | 401 | authentication_error | Missing or invalid API key. | Check the header; make a new key if in doubt. | | 402 | insufficient_quota | Not enough credits. This request needs about N. ... | Add credits. N is the worst case for this request, so a smaller max_tokens can get it through. | | 404 | invalid_request_error | The model 'x' does not exist. | Ids are exact. Copy from the catalog. | | 429 | rate_limit_error | This key's daily cap of N credits would be exceeded. ... | Raise the cap in the wallet or wait for the window to roll. | | 429 | (plain body) | {"statusCode":429,"error":"Too Many Requests","message":"Rate limit exceeded, retry in 1 minute"} | 120 per minute per IP. This one is not in the OpenAI envelope. Back off and retry. | | 400 | invalid_request_error | web_search is not available on this server. (code W206) | This server has no search. Drop web_search. param is web_search, code is W206. | | 400 | invalid_request_error | This model cannot use web_search; see supportsTools on GET /v1/models. (code W207) | Pick a model with supportsTools: true, or drop web_search. code is W207. | | 429 | rate_limit_error | Search the web is limited to a few turns a minute. Wait a moment and send again. (code W109) | 6 searching calls a minute per account, workspace and API together. Wait and retry. code is W109. | | 400 | invalid_request_error | The model 'x' does not accept image input. | An inline picture on a model whose inputs on GET /v1/models lack image. Pick one that sees pictures. | | 400 | invalid_request_error | The model 'x' does not accept file input. Upload the file with POST /v1/files and send its file_id instead; ... | An inline file, input_audio or video_url part on a model that takes no such input. Upload it and send file_id, and a PDF or sound goes as words. Files. | | 400 | invalid_request_error | Unsupported content part: x. | Only text, image_url, file, input_audio and video_url parts are read. | | 404 | invalid_request_error | No file with id 'x'. | The file_id names nothing on your account. Upload it again. | | 413 | invalid_request_error | At most 6 pdf files per request. | Over the per-request count for that kind. Limits. | | 413 | invalid_request_error | This request, with its files, is about N tokens; the model 'x' takes M. | Over the model's window. code is context_length_exceeded. Pick a bigger window, or send less. | | 400 | invalid_request_error | That model cannot watch video. Pick one that does, or remove the video. | An upload this model cannot take. code carries the workspace code (W609 here; also W203, W605, W606, W607, W608, W611). Error codes. | | 415, 413 or 400 | invalid_request_error | That file type is not accepted. ..., That file is too large. The limit is ..., That file could not be read. ... | An upload refused at POST /v1/files, with W604, W605 or W610 in code. Files. | | 503 | api_error | Model access is temporarily unavailable. Nothing was charged. | Retry shortly. | | 502 | api_error | The model gateway is unavailable. | Retry with backoff. | ## After the model has run | Status | Type | Message | Charged? | What to do | | --- | --- | --- | --- | --- | | 429 | rate_limit_error | The upstream gateway is busy. Retry shortly. | No | Retry after a short delay. | | 400 | invalid_request_error | The model's provider declined this request under its content rules. Asking for a real public figure is a common cause. Nothing was charged. Reword it, or use another model. Its code is content_policy_violation, as on OpenAI. | No | Reword the prompt or use another model. Retrying it unchanged is refused again. | | 502 or 400 | api_error | The model gateway rejected the request. | No | Any other provider failure. 502 when the provider failed, 400 when it turned the request down for a reason other than its content rules. Check the prompt and the model. | | 502 | api_error | The model returned no content. Nothing was charged. | No | A request that returns nothing is an error, not an empty answer. Retry or change the prompt. | | 502 | api_error | The gateway sent an unreadable response. It was billed upstream, so the reserved credits were charged. | Yes, the reservation | Do not retry blindly. Check your activity first. | | 504 | api_error | The model took longer than the gateway deadline. It was generated and billed upstream, so the reserved credits were charged. | Yes, the reservation | Lower max_tokens or pick a faster model. Do not retry blindly. | With stream: true the refusal can also arrive after the stream has started, as an SSE event with the same code. The stream then closes without [DONE], and it is settled like any stream that ends early: on what it carried, which includes the prompt. That is why this message leaves out "Nothing was charged". Any other provider error mid-stream arrives the same way, with type api_error and code upstream_stream_error. ```text data: {"error":{"message":"The model's provider declined this request under its content rules. Asking for a real public figure is a common cause. Reword it, or use another model.","type":"invalid_request_error","param":null,"code":"content_policy_violation"}} ``` > A retry loop that treats every 5xx as free will double-spend on the two charged cases. Retry 503, 502 with unavailable or rejected, and 429. Log 504 and the unreadable 502 and look before retrying. Never retry content_policy_violation unchanged: the same request is refused again. ## Backoff that behaves ```python import time, requests RETRY = {429, 503} def call(payload, tries=4): for attempt in range(tries): r = requests.post("https://heyaskr.ai/v1/chat/completions", json=payload, headers={"Authorization": "Bearer askr_live_..."}, timeout=130) if r.status_code == 200: return r.json() body = r.json().get("error", {}) charged = "reserved credits were charged" in body.get("message", "") if r.status_code in RETRY or (r.status_code == 502 and not charged): time.sleep(2 ** attempt) continue raise RuntimeError(f"{r.status_code}: {body.get('message')}") ``` Source: https://heyaskr.ai/docs/api/errors --- # Limits and caps How fast can I go, how big can a request be, and how do I stop a script overspending? Verified against the code on 2026-09-24. | Limit | Value | Where it bites | | --- | --- | --- | | Requests per minute | 120 per IP address | 429 with no message body change. Back off. | | Request body | 12 MB on chat completions, 100 MB on a file upload, 256 KB elsewhere | 413 | | Uploads per minute | 60 per IP address | 429 | | Files per request | 20 pictures, 6 PDFs, 10 text files, 6 documents, 4 audio files, 2 videos, across every message sent; per message 6, 3, 5, 3, 2 and 1 | 413. Files | | A file | 8 MB for a picture, 32 MB for a PDF or a document, 5 MB for text, 25 MB or 20 minutes for audio, 100 MB or 10 minutes for video | 413 at upload, code W605 | | Conversation | Last 400 messages, up to the model's own context window (about 1 million tokens on the largest). A model with a small window: last 40 messages, 120,000 characters | Older messages dropped; over the limit is 400 | | Output | max_tokens up to 8,192, default 4,096 | Clamped, never rejected | | Upstream deadline | 120 seconds | 504, charged, because the model ran | | Active keys | 20 per account; 10 created per hour | 400 from the wallet | | Daily cap per key | Optional, in credits, rolling 24 hours | 429 rate_limit_error before the call | | Searching calls | 6 per minute per account, workspace and API together | 429 rate_limit_error, code W109, before the call | | Search the web, per call | 3 searches, 2 pages read, 4 model calls | The model is told the limit is reached and answers with what it has. Search the web | ## There is no per-account request limit The IP limit protects the service. The daily cap protects your balance, and it is the one to set. It is checked against the worst case a request could cost before the request runs, so a capped key cannot go over, even by one request. ## Need more Higher IP limits for a server that fans out on behalf of many users: email ask@heyaskr.ai with the address and the shape of the traffic. Source: https://heyaskr.ai/docs/api/limits --- # Differences from OpenAI What is deliberately different from the OpenAI API, and why? Verified against the code on 2026-09-23. The request and response shapes match, so existing clients work. These are the deliberate differences, and there are only five. | Difference | Why | | --- | --- | | askr.credits_charged on every response and on the final stream chunk | Additive, so clients that do not know about it ignore it. Knowing the cost of a call without a second round trip is most of the point. | | 402 when the balance is short | It names how many credits the request needed, so a script can decide whether to top up or back off. | | Chat models that answer with a picture | A model that answers chat completions with an image works here, and the picture comes back in message.images. Video runs in the workspace and is not on this API yet. Audio and embedding ids are refused. | | No fine-tuning, assistants or responses endpoints | Not built. Chat completions, models and files on The API are the whole surface. A file by file_id is routed per model, as the file or as its words, which OpenAI does not do. | | web_search: true on chat completions | The search is ours, not the provider's: the model gets our search and page tools, and the sources come back as OpenAI-shaped url_citation annotations, so a client that reads OpenAI's search results reads ours. Search the web. | ## And the things we quietly ignore temperature, top_p, tools, tool_choice, response_format, stop, n, seed and plugins are accepted and dropped. purpose on a file upload is accepted and ignored. Model id suffixes like :online or :fast are not a feature; they make an unknown id. Chat completions has the full list. Source: https://heyaskr.ai/docs/api/differences --- # Use it with your tools How do I point Cursor, Continue, Aider, the OpenAI SDK or LangChain at askr? Verified against the code on 2026-09-23. Two settings. Anything that speaks the OpenAI API works, because the shape is the one it already sends. - Base URL: https://heyaskr.ai/v1 - API key: askr_live_... from your wallet - Model: Any chat id from the catalog | Tool | How | | --- | --- | | Cursor | Settings, Models: add an OpenAI key, then set the override base URL. Enter an askr model id as a custom model. | | Continue | In config.json: "provider": "openai", "apiBase": "https://heyaskr.ai/v1", "apiKey": "askr_live_...". | | Aider | aider --openai-api-base https://heyaskr.ai/v1 --openai-api-key $ASKR_KEY --model openai/claude-sonnet-5 | | OpenAI SDK (Python, Node) | Pass base_url / baseURL and api_key / apiKey. Nothing else changes. | | LangChain, LlamaIndex | Any OpenAI chat model class that accepts a base URL and key. | | Open WebUI and other chat front ends | Add an OpenAI-compatible connection with the base URL and key. The model picker fills from /v1/models. | | Your own script | It is HTTP and JSON. curl works. | > Claude Code, the Anthropic desktop app and other tools that speak the Anthropic Messages API cannot be pointed at askr: the shape is different. Cursor and Continue can run Claude models through askr because they speak OpenAI's shape. ## What to expect - Streaming works everywhere the tool supports it. - Tool calling and structured output are not applied yet. A tool that depends on function calls will fall back to plain text. Not on the API yet. - Web search is a field on the request, web_search: true, not a tool call. A front end that lets you add request fields can turn it on. Search the web. - A screenshot pasted into a tool goes through as an image_url part, to a model that sees pictures; a PDF a tool attaches goes as a file part. Pictures and files. - Every call from every tool is one line in your activity, with its cost. Source: https://heyaskr.ai/docs/api/tools --- # The CLI How do I talk to any model from a terminal? Verified against the code on 2026-09-15. The terminal client is a single file with no dependencies. It needs Node 18 or newer; node -v tells you what you have. It remembers your key and your model, so you choose once and then type. ## Install ```bash curl -fsSL https://heyaskr.ai/install.sh | sh ``` ```powershell irm https://heyaskr.ai/install.ps1 | iex ``` The installer is deliberately readable and does two things: fetches one file into ~/.askr and puts a launcher on your PATH. Run the line for your machine, not both; Windows has no sh. ## Sign in ```bash askr login ``` Paste a key from https://heyaskr.ai/wallet#api. It is saved to ~/.askr/config.json, so this is once per machine. ASKR_KEY in the environment overrides it. ## Talk ```bash askr > what is a solar eclipse? A solar eclipse happens when the Moon passes between Earth and the Sun. 12 in, 118 out > ▌ ``` askr on its own opens a conversation that remembers what was said until you leave. Answers render as you would want them: bold is bold, lists are lists, code is set apart. Below each answer is what it cost in tokens. | Command | What it does | | --- | --- | | /models | List what you can call | | /model | Switch model without losing the thread; /model alone shows the one in use | | /new | Start a fresh conversation | | /balance | Credits remaining | | /exit | Leave | ## Ask one thing and leave ```bash askr "what is a solar eclipse?" ``` The quotes matter. Without them your shell reads ? and * as filename patterns and stops with no matches found before askr runs. Inside a conversation there is no shell in the way. ## If something does not work | You see | What happened | | --- | --- | | command not found: askr | The installer added a line to your shell config that has not been read yet. Open a new terminal tab, or source ~/.zshrc. | | no matches found | Your shell ate the question. Quote it, or run askr and type it in. | | node: command not found | Node is not installed. Get it from nodejs.org, then run the install line again. | | That key was rejected | The key was revoked, or only part of it was pasted. Make a new one and run askr login again. | | 402, or not enough credits | Your balance ran out. Add credits and carry on. | ## Pointing it elsewhere ASKR_ORIGIN changes the origin the CLI talks to. It defaults to https://heyaskr.ai. Useful for staging and for nothing else. Source: https://heyaskr.ai/docs/api/cli --- # Models Every model on askr, what it costs, and where it runs. Verified against the code on 2026-09-15. This list is the live catalog, read when the page loads and cached for five minutes. Ids are exact and are what you pass as model. Prices are per million tokens for chat, per picture or per clip for media. (The live model catalog is rendered here. JSON: /api/models) ## Reading the table - Runs on: API means POST /v1/chat/completions accepts it. Workspace means the browser can run it. Video is workspace only. - Turn: Credits for a typical chat turn: 1,000 tokens in, 500 out. Thinking models run above this because reasoning is billed as output. - Picture: The measured credits per image where we have measured it; the reservation otherwise. - Clip: The cheapest published shape and length. Longer and larger cost more; the workspace shows the exact price before you render. Ids that the catalog lists but the provider is not currently serving are left out. Listed but refused explains. Source: https://heyaskr.ai/docs/models --- # Choosing a model Which model should I use for what? Verified against the code on 2026-09-15. There is no upgrade to buy and no plan to pick, so the honest answer is: try two. A switch costs one string on the API and one click in the workspace, and the same question on a second model is the fastest way to learn which one you like. | Job | Reach for | Because | | --- | --- | --- | | Everyday questions, drafts, summaries | gemini-3.7-flash | Fast, cheap, a million tokens of context | | Careful writing, editing, code review | claude-sonnet-5 | Strong at long context and following instructions | | An all-rounder with tools in mind | gpt-5.6-sol | The closest match if you are moving OpenAI code | | Hard reasoning where cost does not matter | claude-fable-5.1 | The most capable model on the catalog today | | High volume on a budget | deepseek/deepseek-v4.1-flash | Capable reasoning at a fraction of a credit per turn | | A picture | google/gemini-3.1-flash-image | Fast, cheap, works through chat | | A short video clip | kling-3.0 or veo3-fast | Smooth motion; the workspace shows the price per shape | ## What the numbers mean - Input rate is what you pay per million tokens you send, including the conversation history you resend each turn. Long conversations cost input, not output. - Output rate is per million tokens the model writes, including reasoning you never see on thinking models. - Context is the most the model can hold in one call. Bigger is not better; it is a ceiling. ## Cheap experiments Set max_tokens low while iterating on a prompt. The reservation and the worst case of a runaway reply both shrink with it. Source: https://heyaskr.ai/docs/models/choosing --- # What is not on the API yet Which capabilities are missing today, and where are they on the roadmap? Verified against the code on 2026-09-23. One honest line each, so nobody spends an afternoon on a feature that does not exist. Where an item is on the roadmap, it says so. | Capability | Today | Plan | | --- | --- | --- | | Video on the API | Workspace only. POST /v1/videos does not exist. | A video endpoint is on the roadmap after the router rebuild. | | Audio: speech and transcription | Not available anywhere. Audio ids are refused. | Audio routes are on the back-end roadmap. | | Web search | Built. web_search: true on chat completions, the globe in the workspace. :online and plugins still do nothing. | Live. | | Routing controls | None. :fast, :cheap, :nitro, :floor make an unknown id and return 404. | The new router will expose routing preferences. | | Tool and function calling | tools and tool_choice are dropped; the model answers in text. | On the roadmap. | | Structured output | response_format is dropped. | With tool calling. | | Image input (vision) | image_url parts, inline or by file_id, on models that see pictures. | Shipped. Files. | | Embeddings | No endpoint; embedding ids are refused. | Under consideration. | | Files | /v1/files upload, list, read, delete; file, input_audio and video_url parts. | Shipped. Files. | | Assistants, fine-tuning, responses API | Not built. | Not planned. | | CORS for browser calls | No CORS headers. | Not planned: a key in a browser is a key anyone can spend. | The roadmap has the order and the reasoning. Source: https://heyaskr.ai/docs/models/not-yet --- # Listed but refused Why does a model I can see in the catalog return an error? Verified against the code on 2026-09-15. The upstream catalog lists some ids whose provider is not actually serving them: they answer with an error, or with nothing, or need a route we do not have. Rather than let you find that out with a failed request, askr keeps a list of those ids and refuses them up front. - On the API: Those ids are left out of GET /v1/models and return 404 The model 'x' does not exist. - In the workspace: Not offered in the picker. If one is requested anyway: That model is listed but its provider is not serving it. Pick another. Nothing was charged. - On the catalog page: Not shown. ## Why they are on the list - The model answers 200 with no content and no cost. - The provider returns an error for every request. - No provider will serve the id at all. - The id does not exist upstream any more. - The model needs a private-mode proxy we do not route through. - The model is blocked by the provider's data-policy guardrails. The list is a measured snapshot, re-checked when the catalog changes. If a model you want is missing and you believe it works, tell us and we re-test it. Source: https://heyaskr.ai/docs/models/refused --- # Credits What is a credit, what is it worth, and what are the rules? Verified against the code on 2026-09-20. - One credit: $0.001. One dollar buys 1,000 credits. - What it buys: Model usage, at the model's own rate. Nothing else. - Platform fee: None on usage today. A credit buys the same amount of model work it cost askr. If that changes, the terms page announces it before it applies, and it never applies to credit you have already spent. - Processor fee on deposits: Absorbed by askr today. You are credited for the full amount you paid; the processor's cut, about 1%, is a cost askr carries. If that changes it is announced on the terms page first. Deposits. - Expiry: None. Credit stays until you use it. - Withdrawal: None. Credit is prepaid access to model usage and cannot be converted back to crypto or cash. - Interest or yield: None. Credit earns nothing. ## Why credits and not dollars Model prices are small fractions of a cent. A short chat turn on a cheap model costs a few hundredths of a credit, and a number like 0.032 reads better than $0.000032. Everything on the site, in the wallet and in askr.credits_charged is in credits; the wallet shows the dollar equivalent beside the balance. ## Where a deposit goes The address is issued by the payment processor. Your payment lands in askr's balance there, and from it askr sends the part that backs your credit on to the model gateway that runs the models. Each dollar you paid buys 1,000 credits; the processor's cut comes out of askr's side, not yours. ## When a platform fee is introduced It will be one multiplier on top of upstream cost, shown as such, and holders of the askr token pay less of it on a published scale. It will never touch the upstream price, and the docs and the terms will say the number. Source: https://heyaskr.ai/docs/billing/credits --- # Deposits Which assets, which networks, what limits, and what happens in each case? Verified against the code on 2026-09-20. A deposit is one address, one asset, one amount, one countdown. Credits land when the payment processor confirms the payment. The assets below are read from the processor as it is right now; if it is not listed here, we cannot invoice it. (The live deposit asset list is rendered here. JSON: /api/deposits/assets) ## Limits - Minimum: $5 per deposit - Maximum: $10,000 per deposit - Address validity: Where a countdown is shown next to the address, send within it. An address is for one deposit; make a new one rather than reusing it. - Rate: Fixed when the deposit is created. Not the rate on the page earlier, not the rate at confirmation. ## What happens when | Case | What we do | | --- | --- | | Exact amount arrives | Credited at the quoted dollar value, in full, once the processor marks it complete. | | Less than the quote arrives | Credited pro rata for what arrived. A payment far short of the quote may sit as partial until the processor settles it. | | More than the quote arrives | Credited for all of it at the invoice rate, up to four times the quote. Send in one transaction. | | Arrives after the countdown | May not reach your balance, and we may not be able to recover it. Create a new deposit rather than reusing an old address. | | Wrong asset or wrong network | Usually unrecoverable. Neither you nor we can reverse a chain transaction. | | Sent to an address not shown in your signed-in wallet | Not ours. Never send to an address someone gave you elsewhere. | ## Fees Network fees are yours: send the quoted amount on top of what your wallet charges. The processor takes a small percentage for the payment and the conversion, about 1% at the moment; today askr absorbs it, so $100 sent lands as 100,000 credits. There is no askr fee on top. If askr ever stops absorbing the processor's cut, the terms page says so before it applies. ## Timing The processor confirms when the network does. Fast chains take seconds to a minute. Bitcoin on-chain can take longer at busy times. Your activity shows the deposit as pending until it is credited. ## Privacy A deposit address is issued for that deposit alone and tied to your account so the credit reaches you and nobody else. We keep the deposit record; we do not need or ask for anything about the wallet you sent from. Something wrong with a deposit: email ask@heyaskr.ai with the deposit id from your activity and the transaction hash. Source: https://heyaskr.ai/docs/billing/deposits --- # Holds and settlement Why is my available balance lower than my balance, and when does it come back? Verified against the code on 2026-09-15. Every request reserves the most it could cost before it runs, then settles to what it actually cost when it finishes. The gap between balance and available is those reservations. 1. Reserve. For a chat turn: estimated input tokens at the input rate plus the output ceiling at the output rate, or the image floor for a picture model. For a clip: the exact published price. 2. Run. If the reservation cannot be made, the request stops with 402 and nothing runs. 3. Settle. The reservation becomes the real charge, using the gateway's own cost figure for the call. If the real cost is somehow above the reservation, the charge is clamped to what you have. ## How long a hold lasts - Workspace chat: Up to 2 minutes - API chat completions: Up to 5 minutes - Video: Up to 20 minutes A request that finishes settles immediately. A request that dies without settling is swept within a minute of its hold expiring, and the credits return to available. You never lose credits to a hold. ## Reading it in the wallet Balance is the ledger. Available is balance minus open holds. If they differ and nothing is running, wait a minute and refresh. Source: https://heyaskr.ai/docs/billing/holds --- # What you are charged when Which outcomes cost credits, and which do not? Verified against the code on 2026-09-23. The rule is simple: you pay when the model ran. The table is the rule applied to every outcome. | Outcome | Charged | | --- | --- | | Reply completed | The real cost of the tokens (or the picture, or the clip) | | Stream stopped by you part-way | What was generated up to that point | | Reply came back empty | Nothing. An empty reply is an error, not an answer. | | The provider declined it under its content rules (400 content_policy_violation, W415, W816 for video) | Nothing. Through the API with stream: true it can instead arrive mid-stream, and that stream is settled on what it carried, like one you stop. | | The model gateway rejected the request for another reason (502 rejected) | Nothing | | Model access unavailable (503) | Nothing | | Too many requests (429) | Nothing | | Not enough credits (402) | Nothing. The request never ran. | | Model ran but the reply was unreadable (502 unreadable) | The reservation. It was billed upstream. | | Model ran past the 120 s deadline (504) | The reservation. It was generated and billed upstream. | | Video render failed | Nothing. The hold is released. | | Video render completed | The published price shown before you started | > The two charged failures are rare and both mean the model did the work. Lowering max_tokens shrinks what the reservation can be. Source: https://heyaskr.ai/docs/billing/charges --- # Daily caps How do I stop a script from spending my whole balance? Verified against the code on 2026-09-15. A daily cap is a number of credits one API key may spend in any rolling 24 hours. Set it when you make the key, or on an existing key in the wallet. It is optional, and it is the one setting every script key should have. ## How it is enforced - Before each request, the key's spend over the last 24 hours is added to the worst case this request could cost. If that would exceed the cap, the request is refused with 429 rate_limit_error and does not run. - Because it checks the worst case, a capped key cannot go over the cap, even by one request. - The window rolls: spend from 25 hours ago no longer counts. ```json { "error": { "message": "This key's daily cap of 500 credits would be exceeded. It has spent 488.2 in the last 24 hours.", "type": "rate_limit_error", "param": null, "code": null } } ``` ## What it does not cover Caps are per key. The workspace has no cap; a person typing cannot run away the way a loop can. There is no account-level cap today. Source: https://heyaskr.ai/docs/billing/caps --- # Refunds and withdrawals Can I get credits back out, and what if a figure is wrong? Verified against the code on 2026-09-24. - Withdrawing credits: Not possible. Credit is prepaid access to model usage and is never redeemable for money or crypto. - Chargebacks: None on crypto: crypto transfers are irreversible. On a card payment, a chargeback for credit you used breaks the terms and askr contests it. Pay by card. - A wrong charge: Email ask@heyaskr.ai with the activity line. If we charged you for something the model did not deliver, we put it right. - A wrong figure on the site: Prices and conversions are indicative and move with the market and the catalog. Send the page URL to ask@heyaskr.ai and we correct it. The rules in full are in the terms, summarised on Terms, in short. Source: https://heyaskr.ai/docs/billing/refunds --- # Referrals What is my @, what does a referral pay, and when? Verified against the code on 2026-09-26. Every account has an @. It is your name on the site and your referral code: anyone who joins through your link is yours, and what they top up pays you, in credits, for as long as they use askr. Since v1.2, 20 September 2026. - Your @: A username account's handle is its @ from the start. An email or wallet account picks one on the profile tab of your account: 3 to 32 letters, digits or underscores, no password needed. A handful of words are reserved (askr, admin, support and the like). - Your link: heyaskr.ai/r/. It lands on sign-in with your code attached and leaves a cookie for 30 days, so someone who reads about askr today and signs up next week is still yours. First touch wins: a second referral link does not replace the first. A public collab you made is a link too: its Sign in carries the room, and the first public room a signed-out visitor reads is kept in a cookie for 30 days. - What counts as yours: Any account created while your code is attached, through any door: username, wallet or email. Also any new account made from a public collab you created, unless it came with someone's code, which counts first. A room's referral is the same referral as your link, at the same rates and nothing more. A private room refers nobody, and an account that already exists is never moved. The link is written once, at sign-up, and never rewritten. You cannot refer yourself. - Top-ups: 10% of every deposit a referred account makes, in credits, the moment their deposit credits. $50 topped up by someone you referred is 5,000 credits to you. Automatic, once per deposit, and never out of their balance: they are credited in full. - Fees: 10% of the platform fee on every request they run, once a platform fee exists. askr charges no fee today, so nothing accrues yet; the moment a fee is introduced this starts on its own. Accrued per run and paid to your balance hourly as one line, from 1 credit. - Ceiling: None. As many people as you bring, for as long as they use askr. - Where to see it: The referral card on your profile: your @, your link, how many joined, and what they have paid you. Each payout is a line in your activity with the reason. ## Credits, not cash A referral pays model usage on askr, which is what a credit already is. It is not money, it cannot be withdrawn, and it is not a share of anything: it is a fixed cut of what the people you brought top up and, later, of the fee on what they run. The full rules are in the terms. ## Fair use - One @ per account, and it is public: it appears on your link and is how people find you. - Referring accounts you control yourself, or accounts made to farm the bonus, is abuse; the bonus is reversed and the accounts closed. - A refunded or reversed deposit reverses the bonus it paid. Source: https://heyaskr.ai/docs/billing/referrals --- # The askr token What ASKR is, what holding it does, and what it is not. Verified against the code on 2026-09-21. > Not financial advice, and not an offer. Nothing on these pages is an offer to sell, or a solicitation to buy, any token or security, and it is not investment, financial, legal or tax advice. A token carries risk, including the risk of losing everything you put in. Do your own research and consult a qualified professional before making financial decisions. ASKR is the askr community token. It is access, not a revenue claim: holding it lowers the askr fee you pay and unlocks access bands on the platform. It does not pay yield, it does not share revenue, and nobody earns anything by holding it. - Name: heyAskr - Symbol: ASKR - Chain: Robinhood Chain (chain id 4663), an Arbitrum-based L2 that settles to Ethereum - Venue: Pons v2, paired with ETH. Graduates into a permanently locked Uniswap v4 pool. - Supply: 1,000,000,000, fixed, no mint. All of it starts on the bonding curve; the team gets nothing for free. - Launch: Friday 18 September 2026. The hour is not announced in advance; the links go live here and on X the moment it opens. - Company holding: 5.46%, 54,586,381 ASKR, bought at launch as the first trade and locked. Tokenomics - Utility: 95% off every model, on a daily allowance, plus access bands on the same scale. Holding ASKR - Launch: Where and when it goes live, how to buy on Pons, what a trade costs, and the addresses once they exist. - Tokenomics: Supply, the bonding curve, graduation, trade fees, buybacks, and the company's locked 5.46%. - Holding ASKR: The 95% discount, the daily allowance, the access bands, and how a linked wallet is measured. - Token FAQ: Yield, risk, listings, what happens after graduation. ## The one-paragraph version askr already runs on crypto: you fund a balance with it and spend it on models. ASKR is the piece that lets the people who use askr most get more out of it. Hold it, link the wallet, and every model on askr costs you 5% of its list price, on a daily allowance that scales with how much you hold. Along the same scale come day-one access to new models, priority routing, and a direct line into the roadmap. The discount is on the model price itself, not on a fee: askr is feeless, so 95% of what a model costs is simply taken off your bill. Holding ASKR. ## What it is not - Not yield, not dividends, not revenue share, not a claim on anything askr earns. - Not required. Every model and every feature on these docs works with no token at all. - Not credits. ASKR does not fund your balance; credits do, and they are bought with any asset the wallet accepts. - Not a pre-sale. Nothing was sold to anyone before the public curve opened. The company bought its 5.46% on the same curve as everyone else, in the first trade. Official links: x.com/heyaskr and heyaskr.ai. The token's own record on Pons carries only those two, and they cannot be changed after creation. Anything else claiming to be ASKR is not. Source: https://heyaskr.ai/docs/token --- # Launch Where and when ASKR goes live, how to buy it, and what a trade costs. Verified against the code on 2026-09-21. > Not financial advice, and not an offer. Nothing on these pages is an offer to sell, or a solicitation to buy, any token or security, and it is not investment, financial, legal or tax advice. A token carries risk, including the risk of losing everything you put in. Do your own research and consult a qualified professional before making financial decisions. - When: Friday 18 September 2026. The hour is not announced in advance. Watch x.com/heyaskr and this page: the buy link appears the moment the curve opens. - Where: Pons v2, the launchpad on Robinhood Chain. The token trades against ETH on a bonding curve until it graduates. - Chain: Robinhood Chain, chain id 4663. Explorer: robinhoodchain.blockscout.com. - Addresses: ASKR token contract: 0xa92768863a55d8a0591709f7f5e594a249d36ea3 on Robinhood Chain, verify it on Blockscout. This is the only ASKR contract; anything else claiming to be one is not ours. The curve and pool addresses are published here and on x.com/heyaskr as they go live. ## How to buy on launch day 1. Get ETH onto Robinhood Chain. You need ETH on chain 4663 for the buy and for gas. Bridge from Ethereum or Arbitrum using the chain's official bridge, or withdraw directly from an exchange that supports the chain. 2. Open the ASKR page on Pons. Use the link from x.com/heyaskr, and check that the token address on the page matches the one we publish. The name is heyAskr, the symbol is ASKR, the logo is the red chevron. 3. Buy on the curve. Enter an ETH amount and confirm. The price rises along the curve as more is bought. There are no wallet caps and no buy caps. 4. Hold it in a wallet you control. Holding is what unlocks the fee discount and the access bands, measured from a wallet you link to your askr account later. Holding ASKR. ## What a trade costs | Fee | Rate | Where it goes | | --- | --- | --- | | Pons base fee | 1% | 70% to askr, 30% to Pons | | askr creator tax | 2% | askr | | Total per trade | 3% | Charged on every buy and sell, on the curve and after graduation | Fees are paid in ETH to askr's deployer wallet. They are not shared with holders, and they never will be. The fees fund the company; the token is access. ## Snipe protection Pons applies a 99% tax on trades in the first three seconds after the token is created, so bots cannot front-run the opening. The only exempt trade is askr's own first buy, which is how the company's 5% is acquired. Wait a few seconds after the launch post and buy normally. ## Graduation The curve opens at about 1.68 ETH of value and graduates automatically once 4.2 ETH has been bought. At that point 714 million ASKR has sold on the curve and the remaining 286 million seeds a Uniswap v4 pool on Robinhood Chain whose liquidity is locked for good. Nobody, including askr and Pons, can remove it. After graduation ASKR trades on Uniswap like any token. ## On launch day, in order 1. askr creates the token and makes the first buy in one transaction. 2. The token and curve addresses are confirmed on Blockscout and published on X and here. 3. The company's 54,586,381 ASKR is locked in Team Finance and the lock link is published. 4. Trading is open to everyone. 5. Screener and explorer profiles are claimed over the following hours and days. Token FAQ. Source: https://heyaskr.ai/docs/token/launch --- # Tokenomics Supply, the curve, fees, buybacks, and the company's locked holding. Verified against the code on 2026-09-21. > Not financial advice, and not an offer. Nothing on these pages is an offer to sell, or a solicitation to buy, any token or security, and it is not investment, financial, legal or tax advice. A token carries risk, including the risk of losing everything you put in. Do your own research and consult a qualified professional before making financial decisions. - Total supply: 1,000,000,000 ASKR, fixed. No mint function, no inflation. - At creation: All 1,000,000,000 go onto the Pons bonding curve. No team allocation, no advisor allocation, no pre-sale, no KOL round. - On the curve: 714,300,000 sell on the curve before graduation. - Pool seed: 285,700,000 seed the Uniswap v4 pool at graduation, with liquidity locked permanently. - Company holding: 54,586,381.5419 (5.46%), bought on the curve as the first trade and locked. Below. - Trade fees: 3% per trade: 1% Pons base fee (70% to askr, 30% to Pons) plus 2% askr creator tax. Paid in ETH to askr. - Buybacks: Off at launch. If turned on later, Pons buys ASKR from askr's fee share and vests it back to the company over five years. - Holder fee-sharing: Never. Fees fund the company; the token is access. ## The company's 5.46%, and the lock askr bought 54,586,381.5419 ASKR as the first trade at launch, on the same curve as everyone else, and locked all of it in Team Finance on Robinhood Chain as one Investor plan: start 19 September 2026, a 6-month cliff with 0% released at the cliff, then 22 equal steps every 30 days. Team Finance counts a month as 30 days. | Period | Unlocks | | --- | --- | | 19 Sep 2026 to 16 Apr 2027 | Nothing. A hard lock, 0% at the cliff. | | 17 Apr 2027 onward | About 2,481,199 ASKR (0.248% of supply) every 30 days, 22 times | | 6 Jan 2029 | Last release. About two years and four months after launch. | The lock is on chain and the link is published with the launch announcement and on Launch. The locked tokens and the deployer wallet get no allowance and count toward no band. ## The curve, in numbers - Opens at: About 1.68 ETH of value, roughly a $4,200 fully diluted valuation at launch-week ETH prices - Graduates after: 4.2 ETH of buys, roughly a $51,000 fully diluted valuation - Caps: None. No per-wallet cap, no per-buy cap. - After graduation: A Uniswap v4 pool on Robinhood Chain, liquidity locked for good Dollar figures depend on the ETH price on the day and are illustrative. The ETH figures are the curve's parameters. ## Circulating supply Total supply minus the 54,586,381 locked by the company, minus whatever is still unsold on the curve until graduation. After graduation, total supply minus the locked amount, which then releases about 2,481,199 a month from 17 April 2027. ## What can change later Almost nothing. The name, symbol, supply, fees and socials are fixed at creation. The only two settings that can change are the wallet that receives askr's fee share and the buyback switch. Source: https://heyaskr.ai/docs/token/tokenomics --- # Holding ASKR What holding the token does for you, and how it is measured. Verified against the code on 2026-09-22. > Not financial advice, and not an offer. Nothing on these pages is an offer to sell, or a solicitation to buy, any token or security, and it is not investment, financial, legal or tax advice. A token carries risk, including the risk of losing everything you put in. Do your own research and consult a qualified professional before making financial decisions. Two things, both on one scale, both based on the share of total supply a linked wallet holds. Holding is enough: there is no staking, nothing to lock on your side, and nothing to claim. ## 95% off the models One rule: every model costs you 5% of its list price, up to a daily allowance that scales with your share of supply. For every 0.01% you hold, 1,000 credits of AI a day are charged at 5%, capped at 25,000 credits a day. Past the allowance you pay the ordinary price. It resets at 00:00 UTC, and what you do not use does not roll over. The discount is for credits you bought: airdropped credits and credits given to you are spent first, at the ordinary price, and the holder price starts when your bought credits do. | Held | Share of supply | AI a day | What it costs you | | --- | --- | --- | --- | | 100,000 | 0.01% | 1,000 credits | $0.05 instead of $1 | | 670,000 | 0.067% | 6,700 credits | $0.34 instead of $6.70 | | 1,000,000 | 0.1% | 10,000 credits | $0.50 instead of $10 | | 2,500,000 | 0.25% | 25,000 credits | $1.25 instead of $25 | | 10,000,000 | 1% | 25,000 credits: the cap | $1.25 instead of $25 | > You still need credits to spend. The discount makes the AI cheap, it does not make it free: credits you bought are what the 5% is charged against, and a deposit starts at $5. The allowance is the size of the discounted lane, not a grant. Credits. ## Access bands | Share of supply | Held | Unlocks | | --- | --- | --- | | 0.1% | 1,000,000 | Holder status on the account, early feature flags | | 0.5% | 5,000,000 | New models on day one, a 14-day head start over non-holders | | 1% | 10,000,000 | Priority routing and higher rate limits | | 2% | 20,000,000 | Everything above, plus roadmap input and a direct line | | 4% | 40,000,000 | Everything above, at the largest daily allowance | The scale is continuous for the allowance and stepped for the bands. There are no tiers to argue about: the number is your share of supply. ## How it is measured 1. Link one wallet to your account. From the wallet page, under ASKR Connect. There are two ways in, and both write the same link: Connect wallet, where you sign a message with a wallet that can sign, or Verify holdings, where you paste the address and buy any amount up to 25,000 ASKR from that wallet inside a minute. One wallet per account either way. Both ways are described below. 2. The balance is read on chain, once a day. At 00:00 UTC your ASKR balance on Robinhood Chain is read, and that single reading sets the allowance for the day. Reading once is what makes it fair: at that instant a bag of tokens sits in exactly one wallet, so it gets exactly one allowance however far it travels afterwards. A wallet you link today gets its first allowance at the next 00:00 UTC. 3. The discount applies automatically. Every run is priced against what is left of the allowance before it is charged, so there is nothing to claim and nothing to switch on. Your balance, your discount and the credits left at 5% are shown on your account, in the workspace and in a collab room. - The company's locked 54,586,381 and the deployer wallet never count. - Only ASKR on Robinhood Chain counts. Wrapped or bridged copies elsewhere do not. - Wallet linking is live. The access bands are still being built; the allowance is not, and applies from the next reset after you link. ## Two ways to link a wallet Connect wallet. Sign one message in a browser wallet such as MetaMask or Rabby, or over WalletConnect from a wallet app. No funds move and no transaction is sent; the signature proves the wallet is yours. This is the way in for any wallet that can sign. Robinhood Wallet users can open heyaskr.ai in the wallet's browser tab and tap Browser wallet, or use WalletConnect. Verify holdings. For a wallet that cannot sign a message, such as the wallet inside FOMO, and for anyone who would rather not connect a wallet at all. Paste the wallet address (in FOMO it is under Account, then Deposit, then Deposit crypto: copy the one that starts with 0x); we show what it holds and you confirm it is yours. Then you have sixty seconds to buy any amount up to 25,000 ASKR from that wallet, in whatever app you use, and tap Done. We read the chain, and if the buy was authorised by that wallet the link is written. There is no number to match: what proves the wallet is yours is that the wallet itself made the buy, and nobody else can do that for you. - What to buy: any amount you like, from half a token up to 25,000 ASKR (the page shows the current limit; a bigger buy is an ordinary trade, not a proof). There is no target to hit, so routing, slippage and the token's trade tax cannot make you miss it. - What it costs: whatever you chose to spend, and you keep the ASKR. The only real cost is the trade fee and the gas, so a small buy costs a few cents. - You have sixty seconds from tapping the button for the buy to land, and ten minutes after that to tap Done. A buy made before you tapped the button does not count, nor does one that lands after the sixty seconds; if you missed them, start again. - Nothing to connect and nothing to hand over: no wallet connection, no signature, no keys. The address is only read. - A FOMO buy works. FOMO's wallets are smart accounts that sign their own trades, so the chain records that your wallet made the buy, even though FOMO pays the gas and submits it for you. - A transfer from someone else is refused: anyone can send tokens to any address, so tokens merely arriving prove nothing. If your app delivers buys through a router, the page then gives you sixty seconds to sell what you just bought instead, and that sale is the proof. If another account is verifying the same wallet at the same time and one buy fits both, neither gets it and both start again. - One wallet per account, and a wallet can be linked to one account at a time, whichever way you link it. A wallet verified by a trade behaves exactly like a signed one: the same discount, the same bands, and its first allowance at the next 00:00 UTC. ## Later Two things under consideration once askr has revenue, neither promised: committed model capacity for active API users, and a buyback rule with a published revenue trigger rather than a discretionary one. If either ships, it will be documented here before it applies. Source: https://heyaskr.ai/docs/token/holding --- # Token FAQ The questions holders ask, answered plainly. Verified against the code on 2026-09-22. > Not financial advice, and not an offer. Nothing on these pages is an offer to sell, or a solicitation to buy, any token or security, and it is not investment, financial, legal or tax advice. A token carries risk, including the risk of losing everything you put in. Do your own research and consult a qualified professional before making financial decisions. ## Does ASKR pay anything? No. No yield, no dividends, no share of fees or revenue, ever. Holding it makes the AI itself 95% cheaper for you and unlocks access. That is the whole utility, on purpose: it buys you something to use, never a claim on what askr earns. ## Do I need it to use askr? No. Every model, the workspace, the CLI and the API work without it, at the prices on Pricing. Credits pay for usage; ASKR changes how much of that price you pay. ## Was there a pre-sale or a private round? No. All supply started on the public curve. The company bought 5.46% in the first trade at the opening price and locked it for six months, then 22 monthly unlocks. There was no KOL round. ## What is the risk? - The price can go to zero. There is no floor and no promise about value. - Pons v2 is a third-party launchpad. Its contracts are reported to be unaudited. askr's own exposure is its 5.46% buy. - Robinhood Chain is a new network. Bridges and wallets that support it are fewer than on Ethereum mainnet. - Only the addresses we publish on x.com/heyaskr and here are ours. Copies with the same name are not. ## Can I sell? Yes, on the curve before graduation and on Uniswap after. Every trade pays the 3% fee. Selling drops you out of whatever band your remaining balance no longer qualifies for, at the next daily read. ## Where will it be listed? DexScreener, DEXTools and GeckoTerminal index Robinhood Chain automatically, so the pair appears on its own and askr claims the profiles. CoinGecko and CoinMarketCap take a listing form once there is a live market and holders; they can take days. The Blockscout explorer shows the verified token page. ## What happens at graduation? Once 4.2 ETH has been bought on the curve, the remaining 286 million ASKR and the ETH raised seed a Uniswap v4 pool automatically. That liquidity is locked permanently. Trading continues on Uniswap with the same 3% fee. ## Is there a buyback? Not at launch. Pons supports one that buys from askr's fee share and vests the tokens back to the company over five years. If askr turns it on, the rule and the trigger will be published here first. ## How do I link my wallet? From the wallet tab of your account, under ASKR Connect, in one of two ways. Connect wallet: sign a message with a wallet that can sign, such as MetaMask or a WalletConnect app. Verify holdings: paste the address, then buy any amount up to 25,000 ASKR from that wallet within sixty seconds; it needs no wallet connection, no keys and no particular amount. That is the way in for a wallet inside an app that cannot sign, like FOMO, and a FOMO buy works: FOMO's wallets sign their own trades, so the chain records that the wallet made it. One wallet per account either way. Holding ASKR has the steps. Your first allowance is written at the next 00:00 UTC after you link, from whatever the wallet holds at that moment. Anything else: ask@heyaskr.ai. Source: https://heyaskr.ai/docs/token/faq --- # Where to find us Where do updates, launches and conversations happen? Verified against the code on 2026-09-15. - askr Community: The room, on the site: heyaskr.ai/community. Anyone reads it live; speak with an @ and a linked wallet; ask askr together and the room pays. How the room works. - X: x.com/heyaskr: every announcement lands here first, so it can be quoted. New models, prices, the token, the roadmap. - Telegram: The room: an announcement channel plus a forum group for support, daily rituals and token updates. It opens for the launch on Friday 18 September; the link is posted on X and added here the moment it is live. The contract address is only ever posted there first. - fomo.family: After launch, where holders write theses on the ASKR coin page and strangers reading them find their way to the site or the room. - Support: @askrsupportbot on Telegram: a bug, a sign-in problem, a top-up that has not landed, a token question, the API. Pick a category, describe it, a person replies in the same chat. Deep links: report a bug, a top-up, the token. - Email: ask@heyaskr.ai: partnerships, press, anything the bot is not for - Website: https://heyaskr.ai Four doors, one loop: the site converts, Telegram retains, fomo.family gives people a place to argue for askr in public, and X carries what they say back out. There is no rewards programme, no raid group and no paid shilling. Everything is built on people showing real runs and writing real reasons. ## What we post - New models the day they land on the catalog. - Every change to prices, limits or the API, with a note in the changelog. - Token news: the address, the lock, the listings. Only from x.com/heyaskr and the Telegram room; anything else claiming to be ASKR is not. - A weekly usage recap and the roadmap as it moves. ## Reporting something A wrong figure, a broken page, a model that fails: email ask@heyaskr.ai with the URL or the activity line. A security issue: email the same address with security in the subject and we will reply directly. Source: https://heyaskr.ai/docs/community --- # askr Community One public room for everyone around askr: who can read, who can speak, what @askr costs you, and the rules. Verified against the code on 2026-09-23. askr Community is one public room at heyaskr.ai/community, built on askr Collab and made for phones first. Anyone can read it and watch it live, with no account. To speak you need an account with an @ and a linked wallet. In three of its topics every line is a request to askr, and the room pays for the answer, not you. ## Reading and speaking - Reading: Open the page. Every topic, every line and every picture can be read without signing in, and new lines arrive live. The page asks search engines not to index it, but it is public: anyone with the link reads all of it. - Speaking: An active account with an @ (pick one in Settings) and a wallet linked to it. Any of the three ways to link counts: signing in with a wallet, Connect wallet (sign one message) or Verify holdings (ASKR Connect: a small buy from the wallet). Holding ASKR has the steps. What the wallet holds does not matter unless the team switches on a minimum holding, and the room says so when it does. - Your first day: For the first day after you link a wallet (the window is a room setting) you cannot post links or pictures, and you wait 30 seconds between lines anywhere in the room. Everything else works. - When you cannot speak: The bar under the room says why: sign in, claim an @, link a wallet, muted until a time, or banned. Muted and banned accounts still read. An appeal goes through Support, from that same bar. ## The topics | Topic | What goes there | @askr | | --- | --- | --- | | Announcements | News from the team. Only askr's admins post; everyone reads. The newest one shows as a banner in every other topic, and on /token. | No | | General | Chat. Every line here is chat, even one that says @askr; the room points you to Ask askr instead. | No | | X posts | Links to posts on X, one per line, with up to 280 characters of your own beside it. The room draws the post as a card. Open once the team turns it on. | No | | Buys | askr's own buy feed: every ASKR buy over a set dollar value on Robinhood Chain, as it lands. Nobody posts here; askr writes it. Open once the team turns it on. | Feed | | Ask askr | Every line is a question to askr, as Ask or as Code. | Every line | | Images | Every line is a picture prompt: one to four pictures at a time. | Every line | | Video | Every line is a clip prompt. | Every line | The team sets which topics are open, their order and their slow mode; the page shows what is open. A line carries its writer's @, the time and reactions. Reply to a line from its menu, in the same topic, or in General for an announcement. ## @askr and what it costs you Nothing. The room has its own balance, granted by askr, and every answer, picture and clip is paid from it at the model's price. Your own credits and your holder allowance are never touched. In Ask askr, Images and Video every line runs; in every other topic nothing runs and nothing is charged. The model sees your line and, if you replied to one, the line you replied to, each with its writer's @ in front, plus askr's standing instructions. It does not see the rest of the room. It is told to answer you directly and keep it short, never to give financial advice or trading calls, never to repeat links or addresses from the question, and to send questions about accounts to Support. The caps are room settings the team can change, so this page describes them and the room shows the figures: - The room's day. A cap in credits on what the room spends between one 00:00 UTC and the next. The Today's AI card shows spent against cap. When it is spent, or the balance is empty, askr rests until the next day or the next grant; talking goes on. - Your day. So many @askr runs, pictures and clips each per person per day, with a separate cap on pictures attached to chat lines. The Rules card and the composer show what you have left, and all of it comes back at 00:00 UTC. - The room's clips. A cap on clips for the whole room per day, since a clip is the dearest thing the room makes. - One at a time. One @askr run in flight per person, and a handful for the whole room; past that, the room asks you to try again in a moment. - The off switch. The team can rest askr in the room at any time, and raid mode rests it too. The room's text, picture and video models are the team's choice. The composer names the model, and for Video the price of a clip, before you send. ## Slow mode and floods - Slow mode is a wait between your own lines in one topic, set per topic by the team; General has one from the start, and the Rules card lists them. The send button counts it down. Admins and mods skip slow mode; the daily caps apply to everyone. - Floods. For everyone but the team: one line a second, ten a minute, and the same text again within ten minutes is refused. At most five @s in a line, up to three pictures on a line, and a length limit the room tells you about when you go past it. - Raid mode. When the room is under attack an admin switches it on: 30 seconds between lines for everyone, no new links, askr resting, and wallets linked after it began wait until it is off. The room says when it is on. ## Links and addresses - Links in lines cannot be tapped, on purpose. A person's line is drawn as plain text, so a link in it is words, not a button. The one exception is the name of the support bot, which opens the Support sheet from anywhere in the room. askr's answers are drawn with formatting, and askr is told to write no links of its own. - Refused for everyone but the team: invites to other chats (Telegram, Discord, WhatsApp), link shorteners, hosts that read like askr but are not askr's, links with a name@ before the host, raw IP addresses, and links dressed up with unusual characters to slip past a check. The support bot's own link is allowed. Announcements take only askr's own links, even from the team. - Every 0x address in a line is checked by the server against the one ASKR contract and marked as it or not. The contract and askr's official links are held by the server as the room's own rules; anything else claiming to be ASKR is not. ## Moderation and reports askr's admins and mods are known to the server by their role, never by a name or a picture. What they can do: delete a line; mute for an hour, a day or until lifted; ban, and with a ban delete the person's lines from the last 24 hours; pin up to five lines a topic; set a topic's slow mode. Admins also switch raid mode on and off. Mods cannot act on the team, and nobody acts on their own lines. Every action is written to an audit trail. Deleting a question deletes askr's answer to it too, and a deleted line's pictures stop being served at once. - Report a line: open its menu and choose Report to the mods, then a reason: scam, spam, abuse or something else. The mods see the report; the room does not. One report per person per line, and ten open at a time. - Your own lines: you can delete any line of your own. It leaves the room at once for everyone, with askr's answer to it and its pictures. - Appeal a mute or ban through Support: the bar under the room offers Appeal, which opens the support sheet on the right row. - What gets you muted or banned: scams, impersonating askr or the team, asking anyone for keys or funds, spam, floods and abuse. The rules in plain words are in the terms. ## Support Support sits at the top of the room on every screen size, and in every line's menu. It opens a sheet whose first button goes to @askrsupportbot on Telegram, where a person on the askr team answers; the rows below go straight to something broken, a top-up that has not landed, a question about ASKR, an appeal, or someone messaging you as askr. No Telegram: email ask@heyaskr.ai from the address you want the reply on. Email is slower. > askr never messages you first. Support is only @askrsupportbot, and the Support button is the way in; lookalike names are fakes. Nobody from askr asks for your seed phrase, private key or password, or asks you to send funds or verify a wallet. A DM from "askr", "a mod" or "support" is not us. In the room, the words @askrsupportbot and t.me/askrsupportbot are tap targets that open the same sheet; where they lead is fixed in the code, never read from the line. When a mod or askr answers an account question with "For anything about your account, tap Support at the top of the room. askr never DMs you first.", that is the whole answer: accounts, top-ups, sign-in, mutes and bans are handled by support, not in the room. ## X posts A line in X posts is one link to a post on X (x.com or twitter.com, and the fxtwitter, vxtwitter and fixupx mirrors) with up to 280 characters of your own beside it. No other links, no pictures, and the same post not twice within seven days. The topic has its own slow mode. askr's server fetches the post from X once and stores what the card needs: the author's name and @, the text, whether it has a picture or video, its time, and two pictures, the author's avatar and the post's first picture (none at all when X marks the post sensitive). Your browser gets all of it from askr's own address and loads nothing from X unless you open the link. A card viewed a day or more after its last check is checked again; a post deleted, protected or withheld on X is marked gone and its stored text and pictures are removed. Pictures of posts nobody has shared for 90 days are removed. A post by @heyaskr is marked official; a name or @ that reads like askr but is not is flagged as a lookalike. ## Buys askr's own buy feed, written by askr's server; no outside bot is let into the room. It watches a fixed list of ASKR pools on Robinhood Chain and posts each buy over a dollar value the team sets, as it lands: the amount of ASKR and about what it was worth at the time at DexScreener's price, with the buyer's address shortened and a link to the transaction on the explorer. Sells are not posted. When more buys land in a minute than the feed's limit, the rest fold into one line: "And 7 more buys in the last minute, about $1,240 in all." Addresses are public chain data. The feed never links a buy to anyone's @ and never shows a whole address; the explorer does, because the chain is public. If you say in the room that a buy was yours, you have linked your @ to that wallet yourself, in public. Nothing in the feed is advice: the dollar value is the price at that moment and nothing more. ## A private chat on the side Beside the room is a private chat with askr, so the room itself does not have to be one long prompt: Ask askr privately, from the top bar on a phone or beside the room on a desk, or Ask askr about this from a line's menu. Only you see it. It is saved to your History like any chat at /chat, and it is paid from your own balance at the model's price. Holding ASKR lowers the price of every private answer, within your daily allowance; it is not free for anyone. With an empty balance the panel points you to Ask askr, where the room pays. ## What is public - Every line, picture and clip in the room, with its writer's @, the time and the topic, readable by anyone and kept. - Reactions and their counts; who reacted is shown to signed-in viewers only. That you are typing, with your @, to everyone watching. - The room's spend for the day and how many people are watching. - Never: your wallet address, what it holds, your email or your account id. No route in the room returns them. What is kept and for how long is in the privacy notice; the rules of the room are in the terms. If something on this page is wrong, tell us. Source: https://heyaskr.ai/docs/community/chat --- # Getting help Something is not working. What do I check, and who do I ask? Verified against the code on 2026-09-15. | Symptom | Check | | --- | --- | | No sign-in code | Spam folder; wait a minute; five codes per hour per address. Sign in | | 402 or not enough credits | Balance versus available in the wallet; a request reserves its worst case first | | 401 from the API | The key was revoked or truncated. Make a new one. Keys | | 404 model does not exist | Ids are exact; copy from the catalog. Suffixes are not a feature. | | A model errors every time | It may be on the refused list. Pick another. | | 429 from the API | 120 per minute per IP, or the key's daily cap. Limits | | Deposit not credited | Still pending on the network, sent short, or after the countdown. Deposits | | CLI says command not found | Open a new terminal tab. The CLI | | Charged for a failed request | Two failures are charged because the model ran. What you are charged when | ## When you write to us Email ask@heyaskr.ai from the address on the account. Include the time, the model, and for an API call the id from the response or the error message verbatim. For a deposit, the deposit id from your activity and the transaction hash. We do not need your key, ever; if you paste it, revoke it. Source: https://heyaskr.ai/docs/community/help --- # Roadmap Where askr is going, in order. Verified against the code on 2026-09-15. Direction, in order, without dates. The order is what we work on next; it moves when people tell us what they need. The aim does not move: every model, one balance, funded from any wallet, at a price that is honest about what it cost. 1. Launch. ASKR goes live on Robinhood Chain, feeless v1 in the open: every major model behind one sign-in, priced at upstream cost, crypto in and credits out with no card and no subscription. 2. Holder access. Link the wallet that holds ASKR to your account and the account knows what you hold. Bands by share of supply: holder status and early feature flags, new models fourteen days early, priority routing and higher limits, and roadmap input at the top. Nothing to stake, nothing to claim. Holding is enough. 3. askr Skill. Ten skills behind the Ask tile, each a chat with its own screen that builds itself around what you ask and starts before the answer lands: sources with the words that back each sentence, a draft with the changes marked, worked steps revealed one at a time, a Fix card with checked steps, results on bars against your report's ranges, calculators that run on your device, ideas as cards, a side-by-side compare. Shipped in 1.4. 4. Our own payment processor. Deposits without a third party in the middle. Any token on any chain we support lands straight where credits are honoured: no processor fee to absorb, no forwarding step that can fail, more chains and coins added as fast as we can watch them, and settlement we control end to end. 5. askr Mind. What you told one model is available to the next, on your account, under your control. Switch from a coder to a writer to an image model without re-explaining who you are or what you are building. Shipped in 1.5. 6. Router rebuild. Until September every request went through one upstream gateway. We are rebuilding the routing layer to talk to the makers directly, one by one: Claude, GPT and Grok already run on Anthropic's, OpenAI's and xAI's own APIs as the versions marked (direct), at the maker's list price, with the model's thinking shown as it happens and the conversation cached between turns. Still to come: routing tiers so you can ask for fastest or cheapest, zero-retention routing for work that must not be stored, a real image endpoint, speech in and out. Cheaper per request, faster, ours to control. 7. Fees on, and lower than anyone. Only once the router and the processor are ours does a platform fee switch on, because only then is the cost base ours too. The fee will be lower than any comparable platform's and still leave askr a healthy margin, since nobody else takes a cut in the middle. Holders pay less from the first day, per the published rule. 8. askr Pro. A fusion model per person, assembled from your own workload and work type: which models you reach for, what you ask them, what you keep. It routes each task to the model that does it best for you specifically, whether that means the strongest answer or the cheapest one. The most effective model in the catalogue becomes the one built around you. 9. Credit trading. Crypto to credits, credits back to crypto, and credits into a different crypto. The first tradeable AI credit. This changes what a credit is, from prepaid access to something with an exit, and the terms will say so before it ships. 10. Buybacks and burns. When revenue supports it, buybacks switch on and burns are considered, both following a rule published before they start. Never fee-sharing, never yield: ASKR stays an access token. ## Alongside all of it, without pause - New models the day they land on the catalog. - New currencies as they are supported. - Platform fixes every week. - UX changes from what people tell us. - Every one of them recorded in the changelog. Have a request: ask@heyaskr.ai. The roadmap is shaped by what people ask for. Source: https://heyaskr.ai/docs/community/roadmap --- # Changelog What changed, and when? Verified against the code on 2026-09-15. Every release since 1.0, newest first, with what was in it and where it was announced. Below that, what shipped before there were versions. ## 1.61, Direct xAI routing, 2026-09-28 Grok now runs on xAI's own API too. Pick the version marked (direct) under Grok 4.7, Grok 4.6 or five more, and it costs xAI's list price, shows it is thinking from the first second, has twice the room to think, and keeps the conversation cached between turns so a follow-up costs a quarter to read. The gateway versions stay where they were. New features - Grok (direct). Grok 4.7, Grok 4.6, Grok 4.5, Grok 4.3, Grok 4.20, Grok 4.20 Multi-Agent and Grok Build 0.1 straight from xAI, listed under each model as its (direct) version, in chats and in rooms. - Room to think. A direct reply can run to 64,000 tokens, thinking included, and the chat shows the model thinking from the first second instead of a blank wait. Grok 4.7 and 4.6 take the four thinking depths; 4.5 the first three. - Cached conversations. The transcript is cached between turns automatically, so a follow-up reads the history at a quarter of the input price; there is no charge for writing the cache. - List price, to the tick. A direct version costs what xAI charges, and the receipt is xAI's own figure for the call, cache discount included, not an estimate from a rate card. On six of the seven that is 5.5% under the gateway's copy; on Grok 4.7 the gateway's copy is the cheaper one, and the picker lists it first. Improvements - Receipts name the source. A run on a direct version says (direct) on its receipt, and the ledger records where it ran. - Pictures as pictures. A direct version reads a picture you attach as the picture. PDFs and other files go as their text, since xAI does not take a file. - If xAI is busy, out of credit or unreachable, the turn runs on the gateway's version at its price, and the receipt names the one that ran. A refusal never does. Bug fixes Nothing this time. Also - Content rules. The refusal pause covers this lane on its own, as it does the other two: an account xAI refuses five times in a day loses the Grok (direct) versions for a day (W416, on the error codes page), and every other model keeps running. - Privacy. The privacy notice and the docs privacy page name xAI behind the Grok versions marked (direct): it receives the request with a fixed code made from your account id, is asked not to store the conversation, keeps requests up to 30 days to check for abuse and does not train on them. - API. The direct versions are not on the public API; its Grok models run through the gateway as before. ## 1.6, askr Pro, 2026-09-27 A model you never have to pick. A new chat opens on askr Pro, Balanced: it reads each message with the room's memory and the tier, picks the door and the model, starts a skill when a line calls for one, makes the pictures, clips and sounds the line asks for beside the words, packages the lot around what you asked with a preview beside the room, and shows every pick under your line with Change. New features - askr Pro in the picker. The first row of the model picker in a chat or a room, with Cheap, Balanced and Performance beside it. Choose it and askr Pro picks the door and the model for every line at that tier, and the picker's trigger shows the mark and the tier word. The tier is a ceiling and a floor: Cheap never runs a dear model, Performance always picks the best there is, and a hard question moves up to the tier's reasoning models, never out of the tier, whatever the message says. - On from the start. A new chat opens on askr Pro, Balanced, and so does the Ask composer on the home page, whose first send makes a room on it. Pick a model or a skill by hand and that sticks; the tier, or the model you picked instead, is remembered on that device. In a room the creator's pick sets askr Pro for the room and anyone else's holds for their own lines. - What it reads. The message and what is attached to it, the last lines of the thread, the Building and Made lines of the room's memory, the tier, and the doors the room has on. One call to a decision model that answers with probabilities, not words, says how likely each door is, which of thirteen kinds of task the line is, whether it wants the web, how hard it is, whether it carries on the last one and whether it also wants something made; plain code turns the numbers into a door and a model, so nothing you type can talk it out of the tier. The decision has two and a half seconds and runs before the credits are held; when it is slow or down the line still runs, on the room's model through Ask, and the pick says so. - Every pick shown. Under every line askr Pro decided, one line says what ran and why: askr Pro: Claude Sonnet 5 for writing, Balanced; askr Pro kept Claude Sonnet 5, this continues the last answer; askr Pro: Look things up skill on Claude Sonnet 5, Balanced. Change beside it opens the picker; pick a model by hand and the line runs again on it, askr Pro is off for the room from then on, and the model you picked is the room's model. While it works the room says askr Pro is choosing, then which model it picked, and a clip or a sound names the renderer or the voice on its wait row. - No doors to choose. With askr Pro on the door chips under the message box are gone: it picks Ask, Code, Image, Video or Audio for every line itself, and there is no estimate beside Send, since the price is not known before the pick. A model picked by hand brings the chips and the estimate back. A line that opens by naming a picture, a photo of, a portrait of, a logo for, always paints, and a read-aloud ask in any of the common languages is read aloud. - Made beside the words. When a line wants a picture, a clip or a sound as well as words or code, a site and its branding, a plan and its poster, the answering model makes them itself as it writes: it writes each prompt, each one runs through the room's own door on the tier's painter, renderer or voice, every made thing lands under the answer as its own line, and making a picture shows while it does. Nothing sends you to a door to do it yourself. - Packaged around the ask. Code carries the pictures it was made with, by their addresses, instead of a file name you would have to supply. After a run that made things askr Pro checks once, off the turn, whether everything asked for is there and connected; whatever is missing is made in one more run, and a closing card under the last reply sums the package up, Your dog grooming site: 2 pages, logo, hero, gallery, 2 promo clips, 3 blog posts, with the bill for the whole line in credits and about how many dollars, split by clips, pictures, sounds and code or words, a clip still rendering counted when it lands. - The preview takes the room. When a thread has a page to show, on a desktop the rail folds away and the panel takes the right half at full height, Preview first, with a small button that reaches Room settings and People; on a phone one switch moves between the chat and the full-screen preview, with Chat to come back. The page is assembled from the answer's files, its stylesheets and scripts inlined by name, even when the fences carry no names, and every picture, clip and sound made in the thread is put into the page before it is shown. Anchors scroll, a link to another page of the answer shows that page with Back, a link to a made picture, clip or sound opens it in the room, an outside link opens a new tab, and the frame never navigates away. - Files, and Made here. A Files tab beside Preview and Code lists everything the thread made, pictures, clips, sounds and code files, each with Open and Save, a rendering clip saying so. A Made here card in the rail, and at the top of the Room sheet on a phone, shows the newest six pictures and clips, solo or collab, with All opening the Library. - Code is not a wall. In every answer, in rooms, chats and skills, the fences fold into one row, Your code, 3 files, closed until you open it; the files are listed by name with Copy and Save and open one at a time. A short snippet with no file name stays inline. - Skills, without asking. A line that calls for one of the ten skills, looking something up, writing, learning, fixing something, health, translating, money, ideas, a summary or shopping, runs with that skill's instructions and screen in the same thread: the sources, steps, cards or calculators arrive under the answer, and the pick says which skill ran. A thread that already carries a skill keeps its own. - A follow-up keeps the model. When a line carries on the last task, the model that answered last answers again, as long as it is still live and inside the tier; the pick says kept. A new task picks afresh. - Asked before a dear door. A picture or a sound needs a confident read, a clip a more confident one, and a follow-up in a thread whose last line came through that door a lighter one. Under the bar the line goes through Ask with a chip under it, Make it a picture instead, or a clip, or a sound, that runs the same line through that door. Ask and Code need no bar. - Never stopped in your tracks. When a supplier declines a picture, a clip or a sound under its content rules, askr Pro rewords the request once, the brand, company, real person, franchise or copyrighted work replaced by the kind of thing it is, says so in the room, and tries again. A second refusal becomes an offer on your line, what could not be done and what would be done instead, with a chip, Yes, make the sound that way, that runs it. Inside a written answer the model rewords and tries again itself. With askr Pro off a refusal reads as before. Improvements - Search the web, when a line wants it. When the decision says a line wants the web, Search the web is on for that line, under the room's usual rule. - A painter per line in rooms. The Image door in a room takes a painter picked for that one line, from the composer or from askr Pro, without changing the room's model. - A file on the line is read. A line with a document or a data file attached that reads as an unsure Code pick runs through Ask instead, where the model reads the file. - The tier words. The first tier reads Cheap, in the picker, the skill tier pill and the docs. - The picker on a phone. The trigger shows the mark and the tier word alone, so it fits the narrow pill. - Admin. An askr Pro page in the admin area: decisions in the last day and week, fallbacks, kept, changed by hand, media doors picked, average and p95 time, cost, by door, tier and class, the confidence buckets, rooms on askr Pro by tier, and the newest decisions. Bug fixes - In a room, picking a painter for one line answered W201; the line now runs on the painter it named. - A picture call has four times the output room it had, and a picture cut at the budget (W513, four times in a row in one room) is asked for once more on the same hold, so the check after the run no longer has to make them all again. Also - Docs. askr Pro has a page: what it reads before each pick, how it decides, the rules, what it makes beside the words and how it packages it, what happens when a supplier refuses, what it costs and where it is not offered; askr Skills says a skill can start on its own and askr Collab names the picker's first row. - What it costs. Each decision is a fraction of a credit, metered under its own source and charged at zero while askr is feeless, and so is the rewording of a refused request. The model that answers, and every picture, clip and sound it makes, is priced as always; nothing asks before a dear pick, the credits show under the answer as on every answer, and a line that made things shows the whole bill on its closing card. A line that asks for a site, a clip, a song and four pictures at Performance can run to thousands of credits, and the card says what it came to. - Where it is not. A skill thread you opened yourself keeps its tier pill, since the tier there already picks the model; Compare names its own models and never takes askr Pro; the API names a model on every request and does not take it. - Privacy. The decision reads your line the way any model you run does, and the record askr keeps of each decision, for the admin page and for version two, carries no message text. - Measured before it shipped. Over 260 test lines, in six languages, with files, follow-ups and lines that try to name a dearer model, the door was right 98 to 99 times in 100 at every tier, the web read 97 to 98, and a decision took about 280 ms in the middle and under 400 at the 95th, for a fraction of a cent. A room driven through every door, the tier switch, the chips, Change, askr Pro off and the refusals passed 49 checks of 49. - What is next. Version two builds the ladder around you, from what you reach for, ask and keep: the fusion step on the roadmap. ## 1.51, Direct OpenAI routing, 2026-09-27 GPT now runs on OpenAI's own API too. Pick the version marked (direct) under GPT-6 Astra Pro, GPT-6 Sol or five more, and it costs OpenAI's list price, shows it is thinking from the first second, has twice the room to think, and keeps the conversation cached between turns so a follow-up costs a fraction to read. The gateway versions stay where they were. New features - GPT (direct). GPT-6 Astra Pro, GPT-6 Astra, GPT-6 Sol, GPT-6 Luna, GPT-5.3-Codex, GPT-5.4 Mini and GPT-5.4 Nano straight from OpenAI, listed under each model as its (direct) version, in chats and in rooms. - Room to think. A direct reply can run to 64,000 tokens, thinking included, and the chat shows the model thinking from the first second instead of a blank wait. - Cached conversations. The transcript is cached between turns, so a follow-up within half an hour reads the history at a tenth of the input price. Writing the cache on the first turn costs a quarter more on GPT-6; on a long chat the reads take most of the input cost off. - List price. A direct version costs what OpenAI charges, 5.5% under the gateway's copy on every token. Improvements - Receipts name the source. A run on a direct version says (direct) on its receipt, and the ledger records where it ran. - PDFs as files. A direct version reads a PDF as the file itself, its text and its pages, and pictures as pictures. - If OpenAI is busy, out of credit or unreachable, the turn runs on the gateway's version at its price, and the receipt names the one that ran. A refusal never does. Bug fixes Nothing this time. Also - Content rules. The refusal pause now covers each maker's direct versions on their own: an account OpenAI refuses five times in a day loses the GPT (direct) versions for a day (W416, on the error codes page), and Claude (direct) keeps running, and the other way round. Every other model keeps running. - Privacy. The privacy notice and the docs privacy page name OpenAI behind the GPT versions marked (direct): it receives the request with a fixed code made from your account id, is asked not to store the conversation, keeps requests up to 30 days to check for abuse and does not train on them. - API. The direct versions are not on the public API; its GPT models run through the gateway as before. ## 1.5, askr Mind, 2026-09-26 What you told one model is there for the next. Every chat and every room keeps a short memory that goes first into every model's prompt, with one more about you across chats, and all of it is yours to read, pin, edit, forget or switch off. A new chat is now a room of one, the skills and Compare work inside rooms, and a project can hold a team, so a room knows what the other rooms in it are doing. New features - A memory every model shares. After every answer askr Mind notes, in one line each, who is here, what is being built, what was decided, what is open, what was made and what each person prefers. The block goes first into every prompt, every model in a room reads the same one, and the chat says Using askr Mind while a model is being briefed. Nothing is rewritten and nothing is deleted on its own: entries are only added or replaced. - Yours to keep or not. The askr Mind card in Room settings lists what the room remembers, click to edit, Forget on any line, Forget everything with a second press, and Remember this room for the creator. Pin beside Reply puts a line into the memory. After an answer a Mind noted chip says what was just remembered. - Made, with captions, and Prompt used. Every picture, clip and sound made in a chat or a room is listed under Made; a picture gets a one-line caption of what is in it. When you ask a door for something, the prompt the generator gets is written from the memory and your request, and Prompt used under the result opens to the exact words, with the picture it built on. Ask for a made picture, moving, and it goes to the clip as the start frame. - Recall. When a thread holds more than the prompt carries, the words of your request are searched across every line of the room and up to five earlier lines that match are handed to the model, each with who said it. - About you, across chats. One list of who you are, what you build and how you like things, learned from your chats and put in front of every one of them. Settings, Memory shows it, with Remember me across chats, the entries, and Forget everything. A temporary chat, from the dashed bubble in the message box, keeps askr Mind out of that chat entirely. - Projects. One memory shared by the rooms and chats you put in it: make one under Settings, Memory, pick it beside a chat's title or in Room settings, and every prompt in those rooms and chats carries it. Shared brings what askr knows about you into its rooms too; Walled off keeps you out. - Your memory in a room, your choice. Bring my memory into this room, in Room settings, is each member's own switch for that room. A public room never takes it. - Team projects. A project has people in it: invite by @ under Members on the Memory page, up to twelve, and anyone in it can put their private rooms in. The project's memory then says in one line per room what each room is doing, keeps every picture, clip and sound made in any of them with the room it came from, and a question in one room reaches the lines of the others: ask what is happening in the front-end room and the model names the room and the work. A picture asked for in one room can start from a picture made in another. A public room never joins, and leaving takes your rooms out with you. - Recall by meaning. Every room line and memory entry is embedded, so a question about the thing we said about pricing finds it without the word pricing in it, beside the search by word. A long room stays reachable to its first line. - Made from memory, with references. The prompt a picture, clip or sound door receives is written from what the room remembers, and the picture it builds on is handed to the run: a clip starts from the picture, an edit starts from the picture. Prompt used under the result shows the words, and a Prompt from what the room remembers preview appears under the message box while you type in a media door. Improvements - A room of one. A new chat is a room with one person in it, so a chat can do what a room does: threads, New thread from any line to continue from there, Re-run on the newest answer, a model per line from the composer, receipts that say the time taken and the pot that paid, and Export. It pays from your balance as a chat always did, free credits first, then given, then bought with the holder discount, and a run started alone settles alone even if you invite someone while it runs. Recent lists your rooms. - Skills and Compare in rooms. A skill opens as a room whose thread carries the skill: the banner names it, Change starts a new thread under another skill, and the settings live on the thread so everyone in the room sees the same ones. Compare beside the model picker sends one line to two to four models side by side; every run holds its credits before any starts, Pick keeps one, and later lines and the memory read the picked answer only. - Rooms say who spoke and what was made. A model in a room reads each earlier answer with the model that wrote it, and a picture, clip or sound as a line that says which door made it, with its caption. It is told which doors the room has and to say which one makes what you asked for, and never to pass a picture off as its own. - Estimates that match receipts. The price beside Send in a room adds what the memory and a skill's brief put in front of the model, so the number before matches the receipt after. - A room of one reads as a chat. No purse controls until someone is invited, tools and receipts in sentence case, a one-time note under the first answer saying the room remembers, and the project the room is in named on its title line. - Continue in a room. An older chat gets Continue in a room, on the chat and in History: its lines and pictures become a room of one with the memory built at once. - Every door starts a room. The Images, Video and Audio doors on the home page open a room set for that door on the first send, the start frame riding along. - A memory that stays tidy. A near-duplicate of a line already remembered is refused, Who is here says what each person does, a Pick sets the thread's model, clips and sounds get a caption too, and a room folds at most once every twenty seconds however busy it is. - Put this in a project? A new room whose plan resembles one of your projects offers to join it, once. - Admin. An askr Mind page in the admin area: entries by scope and heading, the writer's calls and their cost, the rooms with the most memory. - Handed on in rooms too. A Claude model that thinks through its whole budget in a room and never writes is handed to GPT-6 Sol under its own name, as the chat has done since 1.395, and nothing is charged for the first run. Bug fixes - A room of one short of credits for a picture or a clip now says so as a chat does, with the balance to top up, instead of the room wording. - A sound made in a chat is described to the model as a sound; it used to be described as a picture. - A picture asked for in a young room with the words make them, right after a message about three pictures, reached the image model as those bare words; the prompt writer now reads the last lines even before the room's first memory entry. - An image model that answers a room in words instead of drawing is asked once more, plainly for a picture, on the same hold. - Re-run on an answer ran the wrong line: it now runs the question the answer sits under. - A wrapped caption was cut at its line break, and a writer that ran out of room kept a half-written last line; both now come through whole or not at all. Also - What it costs. askr Mind's own calls are metered under their own source and charged at zero credits while askr is feeless. They never hold a turn up, and a failure in askr Mind never fails a turn. - Docs. askr Mind, askr Collab and askr Skills each have a page. - Privacy. The privacy notice says what askr Mind keeps, that a temporary chat is neither read from nor written to, that your memory enters a room only when you switch it on there and never a public one, and that deleting a chat, a room or your account deletes its memory with it. ## 1.48, ASKR Connect improvements, 2026-09-27 Verify holdings is smoother for apps that buy through a router, as pump.fun does. When a buy like that cannot show whose wallet it is, the sixty seconds to sell start at once, and the sale you make is the proof. New features Nothing this time. Improvements - Straight to the sale. When a buy is turned down, the page shows the sixty seconds to sell right away, with Done, I sold. There is no button to press first, and a sale made after the verdict counts. - The sale step is part of your check. Checks are limited to three an hour, and the sale step used to take one of them. It now belongs to the check you started, so it never uses one up. - A contract says so at once. A sale from an address that is a contract cannot prove it either, so the page says that straight away and points you to Connect wallet, where signing does. Bug fixes Nothing this time. ## 1.47, Model audit, 2026-09-26 A user tested three of our models and said they were older than their labels. We tested all 383, published everything, and shipped what the test taught us. New features - The model audit report. Every model on askr against its label: the user's test, all 383 chat models with their verdicts, what we changed, and how to check for yourself. Improvements - Every model gets today's date. One line in front of every chat and room turn, so a model no longer calls last year's news "not yet happened". The API passes your messages unchanged. - Cannot tell yet on the catalog. The 55 models the audit could not settle say so on their card, with the reason. - Old names off the list. gpt-3.5-turbo-0613, gpt-3.5-turbo-16k and wizardlm-2-8x22b answer like newer models, so they are off until our supplier says what serves them. Eleven models that no longer answer are off too, and the API refuses a delisted id instead of running it. Bug fixes Nothing this time. Also - Refund. Everyone who ran Grok 4.7 got back what its route's hidden prompt tokens cost them, with no need to ask. ## 1.46, Rounding off, 2026-09-26 A tidy-up release. A public collab you share now brings you referrals, only one sound plays at a time, every error code explains itself when you tap it, and GIFs move in collab chats. New features - Public collabs refer. Share a public collab you made, and anyone who makes their askr account from it becomes your referral: 10% of what they top up comes to you in credits, the same as your own link. An account that already exists, a private room, or a sign-up that came with someone's code does not count. Referrals has the rules. - GIFs in collab chats. A GIF sent to a room moves, at its own size, so a live chat can have some fun. Improvements - Every error code explains itself. The code at the end of an error is now a chip: hover it, or tap it on a phone, and it says why it happened and what to do, with a link to its row on the error codes page. - One sound at a time. Play a song, a voice or a clip and whatever else was playing pauses, in chats, in collabs and in your library. Read aloud does the same. Bug fixes - Adding credits to a room, making one or joining one now shows the code when it fails, like every other error. - A collab run or reading short of credits now carries a code (W301, W825). - A reading that is too long (W821) and a start frame that is too large (W605) showed their code twice; they show it once. - Read aloud on a second reply while the first was reading stopped both; the second now takes over. Also - Docs. Every code on the error codes page has its own link, and W914 (the card checkout box) is listed. Referrals covers public collabs. ## 1.45, Voice, music, sound, 2026-09-25 Audio is the fifth door. Have a voice read your words, ask for a song, or describe a sound and get it, in chats and in collabs, priced per character, per song or per second like everything else. New features - Voice. Twelve voices read what you type: ElevenLabs Turbo 2.5 by fal at 50 credits per 1,000 characters, ElevenLabs v3 by fal at 100, and ElevenLabs Flash 2.5 and v3 through the gateway. Up to 5,000 characters a reading (3,000 on v3). - Music. Describe a song and get one with lyrics: Lyria 3.5 by fal at 100 credits a song, Lyria 3 Pro at 80, and ElevenLabs Music by fal at 600 credits a minute for a piece between 30 seconds and five minutes. - Sound. A door creaking, rain on a tent, a bell: ElevenLabs Sound Effects by fal at 2 credits a second, one to 22 seconds. - In collabs too. An Audio pill in the room composer; @askr with the words or the description and the room hears it, paid from the room balance, kept in the room library. Improvements - A player on the line, lyrics under a song, a Made with AI line under every piece, Download on the receipt, and the library keeps your audio beside your pictures and clips. - Priced before you send. The button says what a reading, a song or a sound will cost, and a refused request costs nothing. Bug fixes Nothing this time. Also - Docs. Make audio covers the models, voices, lengths and prices; the codes W820 to W829 are on the error codes page. - Privacy. The privacy notice says what fal receives for a reading, a song or a sound: the text or the description and the name of the voice, kept under fal's terms, used to train nothing. - API. Audio is not on the public API yet. ## 1.44, Public collabs, 2026-09-25 A collab can now be public: share the link and anyone can watch it live, no account needed, and anyone with an account can join, up to a hundred people on one balance. Private stays the default, and the creator now decides what a room can do. New features - Public collabs. On the People card, or when you start a room, pick Anyone with the link. The room, its answers as they stream, its pictures and clips and its library are then open to anyone who has the link, signed in or not, with a Join button for those who are. Invite only puts it back the way it was; whoever joined stays. - Join from the link. A public collab takes anyone with an account and an @, up to 100 people (a private one keeps 12). Past that the link is view-only. Sign in to join brings you straight back to the room. - What a room can do. Room settings has a Can use row: the creator turns Ask, Code, Image and Video on or off for the room, and the composer offers only what is on. - Room settings for everyone. Every member, and every viewer of a public room, sees the settings and the room balance; only the creator changes them or adds credits. Improvements - A shared link looks like the room. A collab link unfurls with the askr mark and the room's title on the card; a private room's card names no title. Every other page of heyaskr shares with the brand card too. - The room's side. Room settings, then Room (the threads, renamed), then People, from the top of the screen and building downwards as settings open, threads are made and people come in. - Copy link beside the choice, and the New room form asks who can see it. Bug fixes - Public, and how many are watching, on the room's title line; the phone's sheet offers Sign in to a reader with no account. Also - Privacy. The privacy notice says what public means for a collab: anyone with the link can read it, watch it live and play what was made in it, without an account. ## 1.43, Direct Anthropic routing, 2026-09-25 Claude now runs on Anthropic's own API. Pick the version marked (direct) under any Claude model and it thinks as long as it needs, shows its reasoning as it goes, costs Anthropic's list price, and keeps the conversation cached between turns so a follow-up costs a fraction to read. The gateway versions stay where they were. New features - Claude (direct). Opus 5, Sonnet 5, Fable 5.1, Opus 5.5 and Haiku 4.5 straight from Anthropic, listed under each Claude model as its (direct) version, in chats and in rooms. - Room to think. A direct reply can run to 64,000 tokens, thinking included, and the reasoning streams while it happens instead of a silent wait. The handoff to another model when Claude ran out of room no longer applies to these versions. - Cached conversations. The transcript is cached between turns, so a follow-up reads the history at a tenth of the input price or less. On a long chat that takes most of the input cost off. - List price. A direct version costs what Anthropic charges, 5.5% under the gateway's copy on every token. Improvements - Receipts name the source. A run on a direct version says (direct) on its receipt, and the ledger records where it ran. - Native PDFs. A direct version reads a PDF of up to a hundred pages as the file itself, and pictures as pictures. - If Anthropic is busy or unreachable, the turn runs on the gateway's version at its price, and the receipt names the one that ran. A refusal never does. Bug fixes Nothing this time. Also - Content rules. An account that a direct version refuses five times in a day loses the direct versions for a day (W416, on the error codes page). Every other model keeps running. - Privacy. The privacy notice and the docs privacy page name Anthropic behind the versions marked (direct): it receives the request with a fixed code made from your account id, keeps it for 30 days under its commercial terms and does not train on it. - API. The direct versions are not on the public API; its Claude models run through the gateway as before. ## 1.42, Video by fal, 2026-09-25 Seven versions of the video models you already use, served by fal at its list price: Kling 3.0 at less than half today's price, Veo 3.1 in place of Veo 3, and Wan 3.0 and Seedance 2.0 at about 13% less. Pick them in the video picker under the model they belong to, labelled "by fal". Same models, lower price, nothing else changes. New features - Kling 3.0 at less than half price. Kling 3.0 Pro by fal renders a 5-second 1080p clip for 840 credits instead of 1,932, and Kling 3.0 Standard for 630 instead of 1,449. The 10 and 15 second clips are priced the same way. - Veo 3.1 is here. Veo 3.1 and Veo 3.1 Fast come in as versions of Veo 3: an 8-second clip with sound for 3,200 credits instead of 3,680, or 1,200 instead of 1,380 on Fast. - Wan 3.0 and Seedance 2.0 by fal. About 13% less than today: a 5-second Wan 3.0 clip for 500 credits instead of 575, Seedance 2.0 for 1,520 instead of 1,740 and Seedance 2.0 Fast for 1,210 instead of 1,390. - Sound on every clip, and a start frame on all but Veo 3.1 Fast, with the same lengths and shapes each model offers today. Improvements - The price you see is the price you pay. A fal render is never charged above the quote the picker showed. If fal bills more, askr absorbs it. - A queued render keeps its hold. fal queues work when it is busy, so a render keeps its credits held for up to two hours while it waits, instead of the hold lapsing at twenty minutes. - If fal cannot take the job, the render runs on today's route at today's price, and the receipt names the model that ran. - Clearer names in the picker. A supplier's newer model sits under the model it replaces by its own name, as Veo 3.1 by fal under Veo 3, and a version that only starts at a longer length says what the model charges for the same clip. Bug fixes Nothing this time. Also - Privacy. The privacy notice and the docs privacy page name fal behind the versions labelled with its name: it receives the request, with your prompt and any start frame, and keeps it for 30 days and the clip for at least 7. It makes no zero-retention promise, so pick the plain version when that matters. ## 1.41, Pay by card, 2026-09-24 Add credits by card, Apple Pay or Google Pay as well as crypto. The price you pick is the price you pay, tax included, and the credits land as soon as the payment clears. Card credits are for your own use: to keep fraud and money laundering out, they cannot be sent, cashed out or turned into crypto. New features - Add credits by card. Top up opens on Crypto with Card one tap away, and Balance & deposits now has Add credits by crypto and Add credits by card. Pick $5 to $100 or your own amount, tick that you want the credits now, and Checkout opens Stripe's secure page. A dollar is 1,000 credits, the same as crypto, and the smallest payment is $5. - The ways to pay where you are. Cards, Apple Pay, Google Pay and Link work everywhere, with local options such as Klarna, Pix and Samsung Pay depending on your location. The price shows in your own currency. - Tax included. The price you pick is what you pay: any VAT or sales tax is inside it, and askr adds no fee. - Credits land at once. Back on askr the balance fills as soon as Stripe confirms the payment, and a confirmation email follows. Card credits get the $ASKR holder discount, like any credits you buy. Improvements - Card credits are for your own use, on purpose. We are against fraud and money laundering, so credits bought by card follow the rules below. They keep a stolen card, or money that needs hiding, from passing through askr. Credits bought with crypto work as before. - They cannot be sent to anyone. Send credits refuses them, so a stolen card cannot be turned into credits in someone else's account. Your balance shows how many of your credits came from card. - They cannot be cashed out or turned into crypto. No credit on askr can be withdrawn, so askr is never a route from card money to cash or coins. - A collab room is not a way round it. Card credits can fund a room you share, and whatever the room hands back returns to you as card credits, still unsendable. - No chargebacks on credit you used. Used credit is not refunded to the card, and a chargeback on it is contested with the account's usage record: the purchase, your agreement at checkout and the requests that spent it. A run that fails is refunded in credits, as always. - A dispute pauses the account. While a card payment is disputed, the account cannot send credits or buy more by card. If the dispute is lost, credits from that payment come back off the balance, and anything already spent is owed before the account can run models again. - No referral bonus on card purchases. Signing people up with stolen cards earns nothing. - askr on Stripe's page. The checkout carries askr's name, logo and colours, and your card statement reads LINK.COM* HEYASKR.AI. Bug fixes Nothing this time. Also - Terms and privacy. Card purchases are sold through Link, Stripe's merchant of record. The terms now cover card purchases, refunds and disputes, and the privacy notice names Stripe and Link. Add credits in the docs covers paying by card. ## 1.4, askr Skill, 2026-09-24 Ten skills behind the Ask tile, each a chat with its own screen that builds itself around what you ask, and starts before the answer lands. The model writes plain words and small pieces of structured data; your device draws them as cards, calculators, steps and sources, checks them, and does the maths. New features - Ask opens a chooser. Tap Ask and pick Ask anything, the plain chat as before, or one of ten skills: Look things up, Write and edit, Learn and homework, Fix it and how-to, Health and fitness, Translate, Money, Brainstorm, Summarise, Shop and compare. Each opens a chat built for that job, on the model that suits it, with its own message box hint and its own settings. - A screen per skill. Beside the chat on a desktop, a sheet on a phone, each skill draws what it made. Look things up shows every page it read as a card with the exact words that back each sentence. Write and edit keeps the draft beside the chat, with the changes marked, a word count and reading age worked out on your device, and one-tap rewrites. Learn and homework reveals a worked solution one step at a time with a hint on each, checks your arithmetic on the device, and marks a quiz there too. Fix it shows a Fix card: what you need before you start, the difficulty, the steps easiest first, a did-this-fix-it check under the current step, and any risky step locked until you have read its warning. Health and fitness draws each test result on a bar against the range printed on your report, builds your appointment list to tick off on the day, and shows a care card that says when to get help. Translate gives a two-part card with tappable words and other choices, a clarifying question when the meaning could go two ways, and photo translation for menus and signs. Money runs loan, mortgage, savings, debt and budget calculators on your device with the maths shown, and reads a statement photo into spending by category and the charges that repeat. Brainstorm deals ideas as cards to keep, drop, combine and build on, with a lens to filter them. Summarise gives an overview card with the quote behind each point at the length you choose. Shop and compare asks what matters first, then puts the options side by side with the better one marked on every row. - It starts before the answer. The moment you send, the panel opens with the skill's real shape, empty and honest: the headings, the rows, the controls, and one line on what is coming. Nothing is invented; the only content is your own words. Ask about a mortgage of 250k at 4.5 percent over 25 years with a 30k deposit and the calculator opens with those numbers already in it, free, while the model is still thinking. Say what is broken and the Fix card appears with its steps to come. Ask a question and the pages show up as askr reads them. The ghosts shimmer while a reply streams and hold still when it ends, and on a phone a slim bar appears at send, pulses while askr works, and opens the sheet. - Price, Balanced or Performance. Sits in the message box where the mode chip was, beside the model it runs. Price runs a cheaper model with short answers, Balanced is the everyday middle, Performance runs the strongest models, asks for thorough checked answers, and in Health, Money and Fix it also searches the web. Tapping a tier switches the model; picking a model by hand moves the tier to match, or to Custom. The choice is remembered for that skill's next chat. There are no cost estimates inside a skill: the receipt under each reply says what the turn cost. - Your settings as pills. Each skill's own settings sit over the message box as one kind of pill: your device and its system, the languages and the tone, the level and whether askr guides or explains, your country and currency, what you need from a purchase, the length and format of a summary. Menus and panels float over the chat instead of pushing it down, and the settings carry to your next chat in that skill. - Free on your device. The calculators open at 0 cr with example numbers. The arithmetic check in Learn, the quiz marking, the word diff in Write, the lens in Brainstorm and the reading level all run on your device, with no model call and nothing charged. - Double-check with another model. In Look things up, Learn, Health, Money and Summarise, one tap asks a model from a different family to look for mistakes in the last answer and say what it would change. Only when you tap it, priced as one turn on that model. - Photo first. Fix it, Health, Translate, Money and Learn put a camera button first: a photo of the label, the report, the menu, the statement or the question. Summarise and Health also take a PDF or a document. - Save or share. Whatever the skill made, the draft, the steps, the appointment list, the shortlist, the comparison, the sources, copies as text, saves as a file, or goes to your phone's share sheet. - A skill offered from plain chat. When a plain chat asks about a loan, a symptom, a translation or a fix, a chip offers to ask it in that skill with the question ready to send. Decided on your device from your words, no model call, easy to dismiss. Improvements - How it works. Each skill's instructions are held on our server, added to every turn of that chat with the date in your time zone, your language and your settings, and no client can write them. The model answers in plain words plus fenced blocks of structured data with a kind, such as steps, draft, labs, calculator or citations. Your device cuts the blocks out of the reply, draws each with the skill's own component inside its own error boundary, so a bad block costs one card and never the chat, and shows the rest as ordinary markdown, with maths typeset. - What is checked, and where. A citation in Look things up must name a page the search actually read for that answer; one that does not is repaired to the real page behind the same number, without its quote, or dropped. A quote in Summarise must be words from the text you pasted. The maths in Money, the arithmetic in Learn and the diff in Write run on your device, never in the model. If your words describe an emergency, Health shows the care card on your device before the reply and tells the model to agree with it. Nothing is trusted just because a model said it. - Live reference data. Health reads the matching MedlinePlus topic and Money the Bank of England Bank Rate, UK inflation, exchange rates and, when you ask, crypto prices, fetched before the answer, cached, with a short deadline, so a slow source never holds a turn up. Fetched text is wrapped as data the model is told to quote, not obey. - Every model, one contract. Because the blocks are plain fenced JSON, every model on askr can run every skill; each skill simply opens on the model that suits it, a reasoning model for Learn and Money, a fast one for Translate and Summarise, the strongest for Performance. - One size everywhere. The skill header is one slim row in every skill, the settings are 30 pixel pills, every label is in the site's own face and sentence case, and everything a skill adds over the message box sits on the message box's own column at any screen width. - Search turns always answer. Before the last call of a search turn, the one with no tools, the model is told in plain words that no more searches can run and to answer from what it has. Without it, one turn in fifteen on a fast model asked for another search and wrote nothing. No extra call, no extra cost, and plain chats get it too. - The message box. The Compare models side by side button is gone from the message box; the Compare page itself remains. The cost of a turn shows once, beside send, or in the model pill on a phone. Bug fixes - A shared chat link opened on the home. A link with a chat in its address, opened in a fresh browser, landed on the home page instead of the chat. The address was read after the first network wait, by which time it had already been cleared. - Reopened search chats. A chat that searches now shows the globe lit when reopened, as it did when it was started. Also - Prompt tests. All ten skills were run at all three tiers against a judge that sees the history, the settings and the sources, and every block those runs produced was drawn through the real screens on a server. The test scripts ship with the code. - Testing on staging. Hannah built the ten skills on an isolated branch and tested them on staging; the polish, the frames and the release went through staging before production the same day. Announced on X. ## 1.395, Time to think, 2026-09-24 The Claude 5 family thinks before it writes, and on askr that thinking used to arrive as silence: the chat gave up after three minutes, then the connection upstream gave up at five. Claude now thinks where askr can see it, the chat says what is happening while you wait, and a brief Claude cannot finish is handed to a model that can. New features - Claude thinks out loud. Claude Opus 5.5, Sonnet 5, Fable 5.1 and Haiku 4.5 used to think in silence: not a byte arrived until the answer began, and after about five minutes of that the connection was closed upstream (code W511). Every Claude call now takes the route that streams the thinking as it happens, so the chat shows thinking from the first seconds and a long think is no longer cut short. - Said before it happens. With a Claude model picked, the composer says it thinks before it writes and a big brief can take several minutes. Once a Claude reply has been thinking for 45 seconds, the card says so in full: that it may run out of room, that nothing is charged if it does, and where the brief goes next. - Handed on, not dropped. A reply, thinking included, may be up to 32,768 tokens. When Claude thinks through that whole budget and never writes (code W514), nothing is charged and the same brief runs again on GPT-6 Sol as its own answer, under its own name, without you asking. A direct line to Anthropic is being built into our router so Claude gets the room it needs. Improvements - Ten minutes to think. A model has ten minutes before its first word and three minutes between words once it is writing; an answer may run for twenty minutes in all. - With Search the web too. The wait after the searches, where the model reads the results and thinks, has the same limits and the same notes. Bug fixes - Stopped after one sentence. With Search the web on, three minutes of quiet was taken for a dead connection and the answer stopped with whatever had arrived. Those turns were charged only for the sentence that arrived, and where an allowance covered them, nothing came off the balance. ## 1.39, File attachments, 2026-09-24 Give any model a file. Pictures, PDFs, text and code, Word, PowerPoint and Excel, audio and video go in from the paperclip, in the app, in Collab rooms and on the API. A model that cannot read a file itself gets its words, so a PDF works on every model. New features - Attach almost anything. The paperclip, a drop or a paste takes pictures (PNG, JPEG, WebP, GIF, up to 8 MB), PDFs (32 MB), text and code files (5 MB), Word, PowerPoint and Excel documents (32 MB), audio (MP3, WAV, M4A, OGG, FLAC, 25 MB or 20 minutes) and video (MP4, MOV, WebM, 100 MB or 10 minutes). Up to 6 pictures, 3 PDFs, 5 text files, 3 documents, 2 audio files and 1 video on a message. Uploading is free; a file costs tokens only on the messages that carry it, and the estimate and the receipt both say so. - PDFs and documents work on every model. A model that reads files gets the file. Every other model gets its words, read once on the way in, so the cheapest model in the picker can read your contract too. A PDF over 100 pages goes as text everywhere. - Audio and video, on the models that take them. 32 chat models hear sound and 66 watch video, as the catalogue stood on 24 September. Audio for any other model goes as a transcript, made once and kept with the file. - A Files page. Everything you have uploaded, for chats and Collab rooms, in one list: what it is, what was read from it, and when. Open, download or delete. A chat's files go when the chat goes; an upload never sent is cleared after seven days. - In Collab rooms. Attach the same kinds in a room. The file is the room's, readable by its members, and stays with the room. Improvements - Pictures at full size for edits and video. A picture for an image model to change, or the start frame of a clip, now goes at up to 4,096 pixels on its longest side and 6 MB, not shrunk for chat. 212 chat models see pictures. - Ask again about a file. Your three newest messages that carry files send them in full; older ones are named in a line, so a PDF from the start of a long chat is not paid for on every reply. Files count toward the model's window, and a turn that does not fit says so before it runs. Bug fixes Nothing this time. Also - For developers: POST /v1/files uploads once, then {"type": "file", "file": {"file_id": ...}} in any message, in OpenAI's shape, so client.files.create works as it is. image_url, file, input_audio and video_url parts go inline too, and inputs on GET /v1/models says what each model takes. Files. - In the docs. Attach files has the limits, the per-model rules and the costs; codes W604 to W612 on Error codes say what a refused file means. ## 1.388, Credits and clips, 2026-09-24 Airdropped credits are spent first, then credits given to you, then the credits you bought, and every receipt says which. Runway Gen-4 no longer offers a 10-second clip the gateway does not deliver, a clip shorter than the one you paid for is charged at the shorter price, and the render confirmation is askr's own. New features Nothing this time. Improvements - Spend order, spelled out. Free (airdropped) credits go first, then credits given to you (a referral bonus, a transfer in, an operator's credit), then the credits you bought. Every receipt and the balance panel say which pot a run came from. - Charged for the clip that arrived. Every finished clip is measured before it is charged. Shorter than the length you paid for, on a model that prices the shorter length, and you pay the shorter price. askr eats the difference. Bug fixes - Runway Gen-4 renders 5 seconds only, for now. The gateway bills the 10-second price and returns a 5-second clip: on 24 September three request shapes all came back at 5.04 seconds. The 10-second option is withdrawn until it works, and everyone charged for it is refunded. - Render it? is askr's own dialog. The browser's confirm box could be silenced by the browser after a few renders in a row, and the Render button then looked dead until a reload. It cannot happen with ours. The same for the prompt when you edit a message that has replies. ## 1.387, Scroll while writing, 2026-09-24 Scrolling up during a fast answer no longer pulls you back to the bottom. The chat follows an answer until you take hold of it, and follows again when you scroll back down. New features Nothing this time. Improvements Nothing this time. Bug fixes - Take hold and it stays. Since 1.383 the chat followed an answer only while you were near the bottom, but a fast model wrote quicker than you could scroll out of that zone, so a flick was undone within a token or two. The chat now stops following the moment you scroll up, however fast the answer arrives, and the answer keeps growing below you until you come back for it. - On a phone too. A flick that left the transcript still moving when your finger lifted used to count as being at the bottom, and the next token dragged you back. The direction you moved decides now, not where you happened to be when you let go. - Back down, it follows again. Scroll to the bottom and the chat follows the answer as before. A new message of yours still goes to the bottom, and so does opening another chat. ## 1.386, Free credits first, 2026-09-24 Airdropped credits are spent before your own balance, on every model. The balance panel in chat now shows exactly that: an answer paid from free credits moves the free line, not your balance. New features Nothing this time. Improvements - Spent this session counts both. The line under the balance adds what the free credits covered to what the balance paid, so it is what the session cost, not only the part you paid for. Bug fixes - Free credits go first, on every model. A turn, a picture or a clip comes out of your airdropped credits until they are gone, and only then out of your balance. That has been the rule since the first airdrop; the panel in chat was the part that got it wrong. - The panel moves the right number. It used to take every charge off your balance on screen and leave the free line alone until you reloaded, so a session on free credits looked like it was spending your own money. Every receipt now says how much the free credits covered, and the panel takes that off the free line and only the rest off the balance. - Read back after anything irregular. A stop, a reload mid-answer or a render checked later re-reads both numbers from the account instead of guessing. ## 1.384, The whole window, 2026-09-24 Every chat on askr, in the workspace, in collabs and on the API, now carries up to 25 times more per turn: up to about 3 million characters, sized to each model's own context window, which is about 1 million tokens on the largest models. New features - 25 times more per turn. askr used to send any model at most 120,000 characters of a conversation, about 40,000 tokens, however much the model could hold. Each turn is now sized to the model's own context window, with room kept for the reply, up to about 3 million characters. 351 of askr's chat models get bigger turns. - Up to 1 million tokens. The largest models, Kimi K3, GPT-6 Sol, Claude Sonnet 5 and DeepSeek V4 among them, can now use their whole window of about 1 million tokens. A model with a 200,000-token window goes from about 40,000 tokens a turn to about 167,000. - Whole documents, one question. One message can now fill the whole turn, where it used to stop at 48,000 characters. Paste a codebase, a book or a stack of reports and ask about all of it at once. Improvements - Long chats keep their start. Up to the last 400 messages are sent to the model, up from 20, so a long conversation no longer forgets how it began. - In collabs too. A room now reads up to 400 messages of its thread and as much as its model's window holds, up from 20 messages and 100,000 characters, and one message can run past the old 24,000 characters. - On the API too. /v1/chat/completions takes the same: up to 400 messages and the model's own window, up from 40 messages and 120,000 characters. - Kept whole. A long chat is saved in full to your account and stays in your history, without pushing your other chats out of the browser. - You pay for what you send. Tokens are priced as before, and a bigger turn holds more credits before it runs. A model with a small window keeps the limits it had. Bug fixes Nothing this time. Also - Tested at full size. A 779,000-character document, about 161,000 tokens and six times the old limit, went to Kimi K3 in one message. The answer came back in 20 seconds, with a detail from the very first line. - In the docs. Workspace limits, API limits and code W106 now say so. ## 1.383, Room to think, 2026-09-23 Long code answers and big build briefs no longer run out of room. The reply ceiling is four times larger, a model that spends it all thinking now says so, a long answer is given the time it needs, and the chat no longer drags you down while it writes. New features Nothing this time. Improvements - Four times the room. A reply may now run to 32,768 tokens, thinking included, up from 8,192. Long code answers arrive whole instead of stopping mid-file. A thin balance gets the largest reply it can hold rather than a refusal, and a model that takes less than the ceiling is retried one size down on its own, so nobody sees a refusal for it. If an answer still reaches the ceiling, the receipt says so and "continue" carries on. - Time for a long answer. The four-minute cap on an answer is gone. As long as the model keeps writing, the answer keeps coming, for up to twenty minutes, so a long file is no longer cut for time either. Bug fixes - A model that thinks itself out of room says so. Some models, the Claude 5 family among them, think before they write. When it used up the whole budget the reply said "did not answer, try again", which sent people round in circles. It now says what happened, with code W514, and suggests a model that thinks less. Credits go back either way. - Read while it writes. The chat follows an answer only while you are at the bottom. Scroll up to re-read and it stays where you put it; scroll back down and it follows again. A new message of yours still goes to the bottom. Also - In the docs. The reply length and code W514 now say 32,768. ## 1.382, Send credits, 2026-09-23 You can send credits to another askr account. Bought credits and credits you were given move; airdropped promo credits and the holder discount stay where they are. New features - Send credits from your balance. In Settings, Balance & deposits, a Send credits button beside Top up: enter an @, an amount, a note, confirm, and both of you get a receipt and a line in your Sends list. - What moves and what does not. Credits you bought or were given move. Airdropped promo credits keep their expiry and never move, and the holder discount follows your wallet. - Limits that keep it safe. Minimum 100 credits, credits deposited in the last 24 hours wait a day, and a six-digit email code for larger sends. - Final, with a safety net. A send is final once confirmed. If one was made by mistake or by someone who was not you, tell us and we can reverse it while the credits are still there. Improvements Nothing this time. Bug fixes Nothing this time. Also - Still spend only. Sent credits buy model usage like any other; nothing here turns credits back into cryptocurrency or cash. The terms say so. See Send credits. ## 1.381, Search the web, 2026-09-23 Any model can now search the web and read pages when it needs live facts. Off by default; a globe in the composer turns it on, and every search is priced before you send. New features - A globe beside attach. The model may search when your question needs something current. The globe glows while a search runs and shows how many ran. Remembered per chat, with a setting to have it on for every new chat. - Priced before you send. The estimate adds the most a turn can spend on searching, 5 credits a search plus the page text at the model's rate. The receipt shows what actually ran, for example 2 SEARCHES, 10 CR. At most 3 searches and 2 pages a turn. - Sources on every answer. Numbered [n] marks in the reply link to a Sources line underneath, so you can check what the model read. - It knows when to look. Ask about today, a price or the latest news with the globe off and the search switches itself on for that chat, the globe glows, and the receipt shows what it cost. Turn the globe off in a chat and it stays off there. - In Collab rooms too. A globe in the room composer, per room. An @askr line that needs live facts searches, the globe glows for everyone in the room, the reply carries its sources, and the searches come off the room balance like the rest of a room's model use. Improvements - Only where it works. The globe is greyed out on models that cannot use tools, with a note saying so. Zero-retention versions of a model search too. Bug fixes Nothing this time. Also - For developers: web_search: true on chat completions, streaming or not, with sources returned as url_citation annotations and the searches in usage; supportsTools on GET /v1/models. - Privacy: with Search the web on, the search terms the model chooses go to Brave Search; pages are read by askr's server, never from your browser. See Search the web. ## 1.38, New chat UI, 2026-09-23 The chat, the collab rooms and everything around them get a new look: the moon and its falling light, a rail that opens into a sidebar, glass surfaces, Geist type, and a light mode beside the dark one. Every result is still priced before you send it. New features - A new look for the whole app. The chat at /chat, every collab room and the pages around them are redrawn: the red moon with its falling light lines on the home, a 64px icon rail that opens into a 264px sidebar, glass chips for the $ASKR price, your balance and your account, and Geist type with every number, price and label in Geist Mono. On a phone: a top bar, a menu drawer, and a composer docked above the keyboard. - Light mode. Choose Light, Dark or System from the sun/moon switch in the rail, the sidebar, the phone menu and the Collab room, from your account menu, or in Settings under Appearance. The app opens in Dark unless you choose otherwise, your choice never flashes the wrong theme, and the site and the docs stay dark. - A home that gets out of the way. The askr logo, then one line, "What are we building?", whose last word rolls through more than seventy words. Five mode tiles in one row: Ask, Code, Images, Video and Collab. The composer is glass, grows as you type, shows the model and its price before you send, and takes attachments, compare and dictation. Recent chats and your Library sit below it. The Audio and Skills tiles, the old welcome line and the footer line are gone. - The model picker. Cmd-K opens a dialog on desktop and a bottom sheet on a phone: search, the pills All, Ask, Code, Image and Video, A to Z and price toggles, keyboard from top to bottom. All lists every model on the platform, Ask the popular and cheap general ones, and Code every text model that can write code, with the code-flavoured ones first. - Top up without leaving the chat. Top up opens as a dialog or a sheet from the balance chip, the sidebar card and the phone balance pill: pick the asset from the ones the processor offers, the amount, and you get the address and a QR that scans in both themes. A deposit you start and close is watched to its end and lands on your balance. - History, Search, Library, Collab and Settings as pages of their own. History groups your chats by day with mode filters, what each chat cost, Rename and Delete. Search covers your chats and your library with highlights. Library shows your pictures and clips as tiles with Download, Re-run, Open chat and Delete. Settings holds your account: General with Appearance and a default model per mode, Balance and deposits, $ASKR benefits, API keys and Privacy. /account now lives at /chat/settings. - Edit forks the chat. Edit a prompt you sent and it goes out again from that point as a new, charged turn, while the old words and the replies they got stay as an earlier version. A control beside the turn flips between versions. - Collab rooms. The room title is a dropdown with a Switch room popover, the chat column sits beside cards for People, Threads and Room settings, and a phone gets sheets for each. Every member sees the room settings and the room name; only the creator changes them. Reactions sit on the bubble they belong to. A room that cannot load says so and offers Try again. "Collab rooms" is now just "Collab". Improvements - Calm inside a chat. Once a conversation starts, in any mode, in a collab room and on the pages, the light lines stop and the moon dims, so nothing moves behind what you are reading. The home keeps its motion. Reduced-motion settings are respected everywhere. - Replies with receipts. Every answer carries the askr mark, the model, the mode and its status, and a receipt line that starts with the price: Copy, Read aloud, Re-run, Compare and Download as plain links. Code blocks have a header with the file name, diff counts and tinted added and removed lines. Pictures show two up, compare shows four cells with Pick, clips show on a card with a rendering row while they render, and Read aloud plays in a small player that fits a phone. - Chats run on their own. Each open chat keeps its address, like /chat?chat=abc, so a reload or a link comes back to it, and /chat on its own is the home. A reply still writing in one chat no longer blocks sending in another, and Stop only stops the chat on screen. Home always opens a fresh chat. - Everything from 1.37, in the new look. Failure cards that say why and whether credits went back, pictures and clips shown whole at their own shape, one-click copy, editing a sent message and deleting from your library are all carried into the new chat and collab rooms, and a reader at the foot of a chat stays there when a picture loads. - Built to be used with a keyboard and a screen reader. Every control is reachable by Tab with a visible red focus ring, dialogs trap focus and close on Escape, the phone drawer makes the page behind it inert, fields show focus on their rounded box, hit areas are at least 44px on phones, and text meets WCAG AA in both themes. Bug fixes Nothing this time. ## 1.375, Second supplier, 2026-09-23 Many models are now served by a second supplier, Engy, at a lower price and with zero data retention. Pick the version you want; nothing else about how askr works changes. New features - The same model, cheaper. In the model picker a model served by both suppliers shows a second row beneath it, "by Engy", with its own price. Pick either. - Eight models to start. The saving is against the same model on askr before, measured on a standard turn. DeepSeek by Engy: V4.1 Flash 86% less, V4 Flash 84% less. GLM by Engy: 5.2 62% less, 5.3 34% less, 5.3 Flash 14% less. Kimi by Engy: K3 38% less. Qwen by Engy: 3.8 27B 90% less, 3.6 35B 71% less. - Zero retention on those rows. Every Engy version except one is served on hardware Engy operates, with prompts and answers never stored or trained on. The row carries a ZERO RETENTION chip and the model button says so. Improvements - Priced the same way as everything else. The estimate beside Send, the receipt after a reply and the holder discount all work as before; the receipt says BY ENGY when that version ran. Bug fixes Nothing this time. Also - For developers: GET /v1/models lists the new versions as engy/ ids, and /v1/chat/completions accepts them, streaming included. ## 1.37, Feedback fixes, 2026-09-23 Fixes and improvements from what people told us. When a model says no, askr now tells you why and that you were not charged, instead of a failure that looks broken. New features - One-click copy on everything. Every prompt and every answer has a Copy, in the workspace and in a collab room: words copy as text, a picture copies as the picture itself where your browser allows, and a clip copies its link. - Edit what you sent. In the workspace, edit a prompt and send it again from that point, after a check that says how many messages it replaces; what those messages made stays in your library. In a collab room, fix the words of your own message, and everyone sees it marked edited. - Delete from your library. Every picture and clip in your library now has a Delete that asks first, so nothing goes with one stray click. A deleted item is gone at once; only the billing record of what it cost stays. - Live prices on the homepage. The card on the homepage now shows the newest Claude, the newest ChatGPT and the most used model from the next most popular brand, each with its price and the $ASKR holder price, which applies within your daily allowance. Improvements - Pictures and videos shown whole. In a chat or a collab room, a generated picture or clip now shows its whole frame at its own shape, sized to fit, instead of a square crop until you click it. Reading further up while one loads no longer pulls you to the bottom. - The same message every time. You see the same reason whether you wait on the page or come back later, and never the provider's raw text. Bug fixes - When a model says no, you hear why. A request a provider refuses under its content rules, like a video of a real public figure, now says so in plain words, tells you nothing was charged, and suggests rewording it or picking another model. Codes W816 for video, W415 for chat and images. - No more "try again" for something that cannot work. A refusal never tells you to retry unchanged, because the same prompt would be refused again. Rules differ by provider, so another model may accept it. - Videos no longer hang on a refusal. A render the provider blocks is marked failed as soon as the provider says so, instead of sitting on "taking longer than expected", and a brief network blip while we check on a render no longer fails one that is still going. - A failed job looks like a failure. In the workspace, a request that fails now shows a failure card with the reason, and says whether your credits went back or nothing was charged, instead of reading like an answer. - Long videos keep their place. A clip still rendering after 7.5 minutes, or through a dropped connection, keeps being watched and lands on its card when it finishes, even after a reload. - In a collab, a clip that fails says so, and its reserved credits go back to the room, and a refusal never tells the room to try again. - Footer: the chain badge, email and copyright now line up on one edge, on desktop and on phones. Also - Every code explained: workspace error codes now lists W415 and W815 to W819. - For developers: the API answers a refusal with 400 and the code content_policy_violation, before or during a stream. See API errors. ## 1.36, ASKR Connect, 2026-09-22 Link the wallet that holds your ASKR without connecting it, signing anything or sending a single token. Paste the address, buy a little ASKR from that wallet in the app you already use, and the chain shows it is yours. Built for FOMO and every app that can buy but cannot connect. New features - Two ways in. Connect wallet signs a message, as before. Verify holdings is new, and needs nothing but your address. - No connection, no keys, nothing sent. Paste your address, press start, and buy any amount up to 25,000 ASKR from that wallet within sixty seconds. Tap Done any time in the next ten minutes and the wallet is linked. - Works from FOMO. A FOMO buy is made by your own FOMO wallet, so it proves the wallet is yours. Not sure of your address? The page shows you where it is: in FOMO, Account, then Deposit, then Deposit crypto. - Only you can do it. Anyone can send tokens to an address, so a balance going up proves nothing; a buy the wallet made itself does. Tokens someone else sends you never count, and neither does a buy made before you pressed start. - The same benefits either way. A wallet linked by a buy gets the same discount and the same access as a signed one, with its first allowance at the next 00:00 UTC. Improvements - Holdings stay fresh. A linked wallet's balance is re-read within ten minutes of any ASKR moving in or out of it. The daily allowance is still set once, at 00:00 UTC. Bug fixes - pump.fun and other apps that buy through a third-party router. A buy that reaches your wallet through a router cannot show the wallet is yours, so the page now asks you to sell the ASKR you just bought, and that sale confirms it. Also - Both ways, step by step: Holding ASKR. ## 1.35, Holder discount, 2026-09-21 Hold ASKR, link the wallet, and every model on askr costs you a twentieth of its price on the credits you bought. How much you get at that price each day scales with how much you hold: 1,000 credits of AI a day for every 0.01% of supply, capped at 25,000. It resets at 00:00 UTC. New features - Every model at 5% of its price, not a fee discount: askr is feeless, so the 95% comes off what the model itself costs and askr covers the rest. - On the credits you bought. The holder price applies to credits you deposited, and the allowance counts only those. Airdropped credits and credits given to you are spent first, at the ordinary price, so a free credit is never doubled by the discount. - Your allowance scales with your holding: 1,000 credits of AI a day for every 0.01% of supply you hold, capped at 25,000 a day. Past the allowance you pay the ordinary price, so nothing ever costs you more than it would without the token. - Read once a day, at 00:00 UTC. At that instant a bag of tokens sits in exactly one wallet, so it earns exactly one allowance wherever it moves afterwards. Splitting a holding across wallets earns the same as holding it whole. - A wallet you link today earns its first allowance at the next reset, which is what stops anyone linking, spending, moving the tokens on and linking again. - Unused allowance does not roll over. It is the size of the discounted lane for the day, not a grant. - Shown everywhere money is: on your balance, in the workspace rail, on the wallet panel and inside a collab room, with the discounted credits left today and the time it resets. It shows at 0 balance too, so you can see what a deposit would buy. - It works inside a collab too: a run in a room is charged against the runner's own allowance first, and the room's purse pays whatever is left. An allowance is per person per day either way, so a room of holders is simply cheaper to run. - On the price list: every rate on /pricing now carries the holder price beside it. Improvements Nothing this time. Bug fixes Nothing this time. Also - You still need credits. The discount makes the AI cheap, not free: a balance is what the 5% is charged against, and a deposit starts at $5. - The docs say the real rule: Holding ASKR replaces the old promise of 25% off a platform fee that was never charged. Announced on X. ## 1.31, Collab fixes, 2026-09-21 What the first rooms needed: pictures and replies and reactions inside a room, code that reads the same in it as out of it, and the joining bug that stopped some invites working at all. New features - Talk as well as prompt: attach pictures, reply to a particular message, and react with an emoji. The picker opens where you pressed it, not in the corner. - Code reads the same everywhere: coloured, labelled by language, one click to copy, inside a room exactly as outside it. Improvements - Your @ is where your picture is, on the profile, instead of buried in the referrals card. Bug fixes - Join actually joins. An invite link could land you on a room you were not in. Fixed the day it was found. - The privacy page says what a room actually does with what is written in it, and what is forwarded to a model. It said three things that were not true of collab. - Crediting an account with no email works again in the admin. Accounts have been able to exist without one since 1.1, and the action still assumed every account had one. Also - Page views, counted: analytics on every page, named in the privacy notice and the policy rather than added quietly. ## 1.3, askr Collab, 2026-09-20 An invite-only room where a few people work the same models together out of one purse. Chat, code, image and video, the same four doors as the workspace, with everyone in the room seeing every answer as it arrives. New features - A room with its own purse. Whoever opens it funds it, and every run in the room is paid by the room. Nobody spends their own balance by being invited. - Invite-only, by link. The room is not listed and not searchable; the link is the door. - All four doors inside it: chat, code, image and video, on the same models as the workspace. Improvements Nothing this time. Bug fixes Nothing this time. Also - A beta, and it says so. Rough edges are known and being worked; tell us what breaks from the box on the roadmap. ## 1.25, Voice and roadmap, 2026-09-20 Speak to askr on any device and it goes straight to the model you chose; every reply can read itself aloud. Both run in your browser, cost nothing and send no audio to askr. The roadmap now lives on the site, with a box to ask for what you want built. New features - Dictation: a mic beside the paperclip in the composer. Tap to listen, tap to stop; sending stops it. Words show as they are guessed and settle as you go, and a pause to think does not end the session. - Read aloud: a link on every finished reply, in your browser's own voice. Code is skipped with a word, links are read as their text, long replies are read in pieces so it finishes. One reading at a time; tap to stop. - See the mic hearing you: four bars beside the mic move with your voice while it listens. - The roadmap, on the site: heyaskr.ai/roadmap shows every version shipped and the plan ahead in order, and has a box that sends what you want built straight to the team. Improvements - The composer fits what you type: the box grows and shrinks with the message, typed or dictated. On a phone the text runs full width with the buttons underneath. - The site asks for the microphone on the first tap, and if you have said no it tells you where the switch is on your platform. Bug fixes - Docs put right: deposits are credited in full (askr absorbs the processor's cut), overpayments are credited, chats live on your account, and referrals have their own page. Also - In your browser, not on our servers: Chrome, Edge and Safari carry the speech engine, so the buttons appear there and stay hidden in Firefox. No audio is sent to askr, nothing is recorded, and neither feature spends credits. The privacy page says so. Announced on X. ## 1.2, Handles and referrals, 2026-09-20 Every account gets an @ and a referral link. Anyone who joins through it pays you for life: 10% of their top-ups in credits the moment they land, and 10% of their fees once fees are on. No ceiling. New features - An @ on every profile: a username account's handle is its @; email and wallet accounts can claim one. heyaskr.ai/r/ is your link. How referrals pay. - 10% of every referred top-up, paid to you in credits the moment their deposit credits. Automatic, once per deposit, never out of their balance. - 10% of their fees, when fees are on. Accrued per run and paid hourly as credits. Nothing accrues while v1 is feeless; switching the fee on is the only change left for that to start. - A referral link survives the sign-up: the code follows every door (username, wallet, email) and holds for 30 days. - Support where things go wrong: @askrsupportbot in the chat rail, under any refunded reply, on the top-up card, in the account tabs and at the foot of sign-in. Improvements Nothing this time. Bug fixes - Wallet fixes: a wallet that is an account's only way in can no longer be unlinked; BNB pays from a wallet with a proper QR; WalletConnect opens one modal, not two; Robinhood Chain is the active chain on phones. Announced on X. ## 1.1, Made for crypto, 2026-09-19 Sign in with a username and password or with a wallet, no email needed. Link the wallet that holds ASKR and the account shows what you hold. Top up the crypto way. New features - Sign in without an email: a username and password, or a wallet signature. An email can be added later for recovery. - Wallet sign-in on any wallet, through WalletConnect as well as browser wallets. - Holdings on your account: ASKR held, share of supply, your band and the fee discount it earns, read from the chain on link, on demand and daily. - Top up the crypto way: ten coins as tiles with a search under them, a checkout card with the address in a box and a Copy button, and a tracker that follows the deposit from sent to credited. - A receipt by email when a deposit credits. - Support on Telegram: @askrsupportbot takes a ticket and relays the answer. - Chats follow the account across devices; the history is named by what each chat asks. - Drop a picture anywhere on the page to attach it; tap a generated image to see it full screen. - The code panel lists files by name, one open at a time, with Save all. - The token page carries the contract address, the Team Finance lock, and the Buy and Chart buttons. Improvements - Models that stop answering hide themselves from the picker instead of failing your request. Bug fixes Nothing this time. Announced on X. ## 1.0, askr is live, 2026-09-19 heyaskr.ai is open: every major model behind one sign-in, priced at upstream cost, paid in crypto with no card and no subscription. ASKR launched on Robinhood Chain the day before. New features - Four doors from one balance: chat, code, image and video. - Feeless v1: a dollar in is a dollar of model; the processor's and the gateway's fees are absorbed. - The developer API at /v1, OpenAI-compatible, with per-key daily caps and askr.credits_charged on every response. - ASKR on Robinhood Chain, launched 18 September, an access token with bands by share of supply. - Docs at home on heyaskr.ai/docs, generated from the code, reading the catalog live. Improvements Nothing this time. Bug fixes Nothing this time. Announced on X. ## Before 1.0 The site as it was built up to launch. ## September 2026 - Docs moved home. These pages replace the old docs site; every old link redirects. The reference is generated from the code and reads the catalog live. - Token launch set. ASKR launches Friday 18 September on Robinhood Chain. The token. - Clips show their first frame while a render runs and after it finishes. - The library keeps what you make. Images and clips are saved with prompt, model and cost, and served to you only. - The workspace counts on the model you chose. Comparing models is an option you take, not something that happens to you. ## August 2026 - Four doors in the workspace: chat, code, image and video, from one balance. - Video in the workspace with the exact price before the render and a hold that is released if it fails. - The developer API at the root of the domain, so a base URL of /v1 works with every OpenAI client unchanged. - The CLI: install, log in, talk. - Per-key daily caps and askr.credits_charged on every response. Source: https://heyaskr.ai/docs/community/changelog --- # Privacy: what we keep What askr stores about me, and what it does not. Verified against the code on 2026-09-28. Short version: we keep what we need to run your account and bill it, and nothing you say to a model. Files you attach (pictures, PDFs, documents, audio, video) are stored on our server so a model can read them and so you can ask about them again. They go to the model's provider for that turn only, are deleted when you delete the chat or the file on the Files page, and are swept after seven days if never sent. Attach files. | Thing | Kept? | Detail | | --- | --- | --- | | Messages and replies | No | A conversation exists in the browser tab you typed it in. The API forwards your messages to the model and keeps none of them. | | Images and clips you make | Yes | In your library, with the prompt, the model and the cost, served to you only. A temporary copy also exists on the model gateway for up to 24 hours. Ask and it is deleted. | | Your email address | Yes | It is the account. | | Sign-in codes and sessions | As hashes | SHA-256 only. A copy of the database contains nothing usable. | | API keys | As hashes | The plain key exists only in your copy. Name, prefix and cap are kept so the wallet can list it. | | Ledger | Yes | Every deposit and every charge: model, tokens, credits, time, and which key made it. This is your activity. | | Deposit records | Yes | Address, asset, amounts, status, and the processor's payment id. Not the wallet you sent from. The processor issues the address and sees the payment; it does not learn who you are. | | Card purchases | Yes | Amount, any tax, credits, Stripe's references, status, refunds and disputes, the IP address the checkout started from, and whether you accepted the terms. Your card details, name and billing address stay with Stripe and Link, Stripe's merchant of record, which sends your receipt. Pay by card. | | Where it runs | | A rented virtual server at OVH in the United Kingdom. Standard request logs (IP address, user agent, URL, time, status) rotate like any web server's. | | IP address | In flight | Used for rate limiting and in server logs, which rotate. Not tied to your account's ledger. | | Anything for training | No | Nothing you type trains a model, ours or anyone's. | ## Who else sees a request The model's provider does, because it answers it. Today requests reach providers through a model gateway; the roadmap replaces that with direct routing. Your account identity does not travel with the request. Because there is no table of chat text on askr's side, a past conversation cannot be sent to you or deleted for you: it was never held. Copies that reached the gateway and the provider follow their retention rules. A version labelled "by" a supplier, such as "Claude Opus 5 (direct)", runs at that supplier instead. Your prompt, the conversation before it, any picture you attach and the reply go to that supplier, and only when you pick its version. No other supplier receives them; where a supplier passes a request on, its line below says so. A version marked ZERO RETENTION comes from a supplier whose terms say it keeps neither the prompt nor the reply. Each supplier, what it runs and what it keeps: - Anthropic (chat). Anthropic's own API, for the Claude versions marked (direct). With each request it also gets a fixed code made from your account id, so it can tell one person's requests apart; never the id itself or your email. It keeps requests for 30 days under its commercial terms and does not train on them. - OpenAI (chat). OpenAI's own API, for the GPT versions marked (direct). With each request it also gets a fixed code made from your account id, so it can tell one person's requests apart; never the id itself or your email. We ask it not to store the conversation; it keeps requests for up to 30 days to check for abuse and does not train on them. - xAI (chat). xAI's own API, for the Grok versions marked (direct). With each request it also gets a fixed code made from your account id, so it can tell one person's requests apart; never the id itself or your email. We ask it not to store the conversation; it keeps requests for up to 30 days to check for abuse and does not train on them. - fal (video and audio). It makes no zero-retention promise: fal keeps the request, with your prompt and any start frame, for 30 days, and the clip for at least 7. For audio it gets the text you have read, or the description of the song or sound, and the name of the voice you picked, under the same terms. For Seedance it also gets a fixed code made from your account id; never the id itself or your email. When Search the web is on, the search terms the model chooses are sent to Brave Search; page fetches are made by askr's server, never from your browser. Neither carries your account. The full policy is at /privacy. Source: https://heyaskr.ai/docs/trust/privacy --- # Security: codes, sessions and keys How are the secrets handled, and what should I do on my side? Verified against the code on 2026-09-15. ## On our side - Sign-in codes are six random digits from a cryptographic generator, hashed together with the email they were sent to, and dead after ten minutes or five wrong guesses. Five codes per address per hour. - Sessions are 256-bit random tokens stored as SHA-256 hashes, in a cookie that is HttpOnly, Secure and SameSite=Lax. Thirty days, and Sign out everywhere kills them all. - API keys are 32 random URL-safe characters after askr_live_, stored as a hash, shown once. - Money is an append-only ledger. The application role can insert and read; it cannot update or delete a row. - Payment webhooks are verified with an HMAC signature in constant time before a single credit moves. - Transport is HTTPS only with HSTS, a strict content security policy, and no third-party scripts on the site. ## On your side - Treat a key like a card number: environment variable, never a repository, never a browser. - Cap every key a script uses. Daily caps. - One key per place, so revoking one costs you one place. - Lost a device: Sign out everywhere in the wallet, then revoke any key that was on it. - Only ever send crypto to an address shown in your signed-in wallet. ## Reporting a vulnerability Email ask@heyaskr.ai with security in the subject. Say what you found and how to reproduce it; we reply directly and fix first, disclose after. Source: https://heyaskr.ai/docs/trust/security --- # Terms, in short The rules that matter day to day, in plain words. Verified against the code on 2026-09-15. This is a summary for people who will not read the terms. The terms win if they differ. - No company name, address or governing law is published yet. The terms say so plainly, and the consumer law of your own country applies. When askr incorporates, both pages change. - Credit buys model usage and nothing else. It does not expire, earns nothing, and is not a claim on askr for money. - A dollar that arrives buys 1,000 credits. The payment processor takes about 1% on the way in; askr charges no fee on usage today and will announce one on the terms page before it applies, never to credit already spent. - What you make is yours. askr claims no rights over model output, keeps your images and clips in your library without promising forever, does not charge for a render that fails, and reruns a lost render rather than paying damages. - Every price, conversion and estimate on the site is indicative. Model prices come from the catalog and can change without notice. The charge on a request is the real number. - Crypto deposits are final. Send the exact quoted amount, on the named network, within the countdown, to the address in your signed-in wallet. Wrong asset or network is usually unrecoverable. - You are responsible for what you ask models to do and for what you do with the output, images and video included. Providers have their own usage policies and a request they refuse is refused here. - Keys are yours to protect. Usage on a key is usage on your account. - The service is offered as is; we work to keep it up and we tell you when something fails, but we do not promise uptime in the first version. The terms carry a change list, so what moved is on the page. - The token is access, not an investment. Nothing about it is advice or an offer. Questions about the terms: ask@heyaskr.ai. Source: https://heyaskr.ai/docs/trust/terms