> ## Documentation Index
> Fetch the complete documentation index at: https://docs.square1.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Introduction

> Square1 is a community-run, OpenAI-compatible LLM gateway

Square1 is an LLM gateway operated for a verified community. It fronts a
pool of frontier and open-weight models behind one OpenAI-compatible API,
so anything that speaks the OpenAI wire format — the official SDKs, RisuAI,
SillyTavern, LangChain, plain `curl` — works by changing the base URL and
the API key.

```text theme={null}
https://inference.square1.dev/v1
```

<Note>
  Access requires a verified community account, and the service is
  region-locked. If you don't have an account yet, start with the
  [Quickstart](/quickstart).
</Note>

## What you get

<CardGroup cols={2}>
  <Card title="Chat completions" icon="message" href="/api-reference">
    OpenAI-compatible `POST /v1/chat/completions` with streaming, tool
    calling, vision input, and a `reasoning_effort` extension.
  </Card>

  <Card title="Embeddings" icon="vector-square" href="/api-reference">
    `POST /v1/embeddings` with fixed 1024-dimension output.
  </Card>

  <Card title="Live model catalog" icon="list" href="/api-reference">
    `GET /v1/models` reflects exactly what your key can call right now,
    including live input ceilings and rate limits.
  </Card>

  <Card title="Quota introspection" icon="gauge" href="/concepts/quota-feedback">
    Every completion carries a quota receipt, and `GET /v1/usage` reports
    your session and weekly standing.
  </Card>
</CardGroup>

## How it differs from OpenAI's API

Square1 keeps the OpenAI wire format but is a shared community pool, so a
few things work differently:

* **Quotas are pooled and weighted.** You draw from a rolling session
  budget and a weekly budget, and each model has a shared daily cap. See
  [Quotas](/concepts/quotas).
* **Responses carry a quota receipt.** The `wellspring` block and a set of
  `x-session-*` / `x-weekly-*` headers tell you where you stand after every
  request. See [Quota feedback](/concepts/quota-feedback).
* **The request schema is a strict subset.** Fields like `n`,
  `presence_penalty`, `logprobs`, `seed`, and `response_format` are not
  accepted. The API Reference lists exactly what is.
* **Model routing is opaque by design.** The API never reveals which
  infrastructure serves a given model — `owned_by` names the model's
  maker, nothing more.

## Bring your own key

If you have your own account with a model provider, you can register that
credential in the dashboard and route requests through it with the
`@provider/model` syntax — see [BYOK](/guides/byok).