Help & privacy

Picking a model

Three models, one subscription. Every one of them runs on hardware we operate, and none of them forward your prompt to a third-party API. Switch per chat from the model menu in the composer; the switch never rewrites a thread you already have.

One honest caveat, because it is the reason people pick us: Moton Sable is the uncensored one — the only model here fine-tuned not to refuse. Fast and Core are ordinary instruct models with the ordinary instincts that come with that, so if you want a difficult question engaged with rather than deflected, pick Sable. Its row in the model menu says “no refusals” in exactly those words.

New chats start on whichever model you set in Settings → Chat — Sable, until you change it.

Your chats are yours

Your prompts and replies are never used to train a model — not ours, not anyone else's. That part has no setting, because it is not optional.

Where a chat is kept is your choice, and the default is the stricter one. Chats live in this browser’s own storage and nothing is written to our database at all, which is what lets us say we don’t store your prompts without an asterisk. The trade is obvious: clear this browser’s data and those threads are gone, and another device can’t see them.

Turning on Save chat history changes exactly one thing — finished exchanges are also written server-side to your account, so the same threads appear on your phone and your laptop. The first time you turn it on it uploads the chats already sitting in this browser, once. Turn it back off and we stop writing; the composer under the text box always states which of the two modes you are in, so you never have to remember.

Deleting is immediate and total on both sides: deleting a chat, or Delete all chats, removes the browser copy and the server copy together, and closing your account takes the rest with it.

The toggle and the current count live in Settings → Data & privacy.

Projects

A project is a name plus one set of instructions that every chat inside it is given — a role, the house style, the things to keep in mind — so you stop retyping the same preamble into every new thread. Those instructions go along as the opening context of each chat in the project; leave them blank and the chats behave exactly like the ones outside.

Start one in Projects with New project, or from a project’s own page with New chat in this project. An existing chat joins through Move to project in the options menu next to its name in the sidebar, and leaves the same way by choosing No project. Deleting a project never deletes its chats — they just go back to the main list.

Keyboard & shortcuts

Three, and only three. The composer is a text box, not a mode, so there is no chord set to learn.

Everything else is on hover: copy a reply, regenerate the last one, edit and resend a prompt you sent. Your draft survives a navigation or an accidental refresh — it is saved per chat in the same browser storage as the chats themselves.

Use your account from Cursor, Cline or an SDK

The same account that powers this chat answers OpenAI-compatible traffic. Make a key in Settings, name it for the tool it is going into, and point that tool at the gateway.

Keys are shown once, in full, the moment you create them — we keep only a hash, so a lost key means a new key. Revoke one at any time from the same screen and it stops working immediately. Make one at Settings → API keys.

Key traffic is billed against your credit balance, not your subscription. The flat plan covers conversations you have here; programmatic calls are metered per token and draw that balance down, so a long agent run in your editor cannot quietly eat the thing you pay for. Spend is attributed to the key that made it, by the name you gave it.

Cursor / Cline settings

base_url = "https://api.moton.io/v1"
api_key = "mt_YOUR_KEY"

Same thing with curl

curl https://api.moton.io/v1/chat/completions \
  -H "Authorization: Bearer mt_YOUR_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "qwen3.8-27b-uncensored",
    "messages": [
      { "role": "user", "content": "Summarise this in two sentences." }
    ]
  }'

model takes a catalogue id — qwen3.8-27b-uncensored is Moton Sable, the one described above — or the name of a deployment of yours. The endpoint is https://api.moton.io/v1/chat/completions, streaming, tool calls and reasoning included, so an OpenAI SDK needs its base URL changed and nothing else.

Plans & billing

One flat monthly price, and messages that are never metered — there is no daily message count, no token allowance, nothing that runs out because you talked a lot.

What a plan does ration is live context: how much of a conversation you can hold in flight at once. That is how the hardware actually costs us — generating tokens is cheap when several people are doing it at the same time, while a long context has to be re-read on every step and sits in memory the whole while. So eight requests at 32k and one at 262k cost you the same budget, which is the fairer of the two ways to measure it.

The rate is identical on every tier — $3.33 of live context per GB per month — so a bigger plan is a bigger bucket plus more requests in flight at once plus priority when the pool is contended, not a better price. If a request would push you past your bucket it waits for room rather than being refused, and nothing is charged for the wait.

Pick or change a plan at Plans. Card details, invoices and cancellation are on Account, which hands off to the billing portal.

Troubleshooting

There are four things that can actually go wrong in here, and none of them are a quota you ran out of.

Console & docs

The chat app is one product. The same account also opens the infrastructure side, where you pay per token instead of per month.