# SaaufAI

> SaaufAI is one gateway to models from many AI companies, with a plain YES/NO disclosure sheet for every model, so you know where your data goes before you send it. It also runs Wisps: always-on AI assistants with their own memory and email address that ask before they act.

SaaufAI's API speaks the OpenAI format: any OpenAI SDK works with the base URL `https://api.saauf.xyz/v1` and a SaaufAI API key (`sk-saauf-...`, from https://ai.saauf.xyz/account/keys) sent as `Authorization: Bearer`. There is no anonymous access; every call needs a key. Prompts and answers sent to the API are not stored, only metadata (time, model, token counts, status). Errors are JSON: `{"error": {"message", "type", "code", "resolution"}}`, where `resolution` says what to do next.

**How to call the API**

1. List the models your key may use: `GET https://api.saauf.xyz/v1/models`.
2. Send a chat completion:

```bash
curl https://api.saauf.xyz/v1/chat/completions \
  -H "Authorization: Bearer $SAAUFAI_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "glm-5.3", "messages": [{"role": "user", "content": "Hello!"}]}'
```

3. Read `choices[0].message.content`. Add `"stream": true` for server-sent events. On `429`, wait for `Retry-After` seconds; on `502`/`503`, retry with backoff or use another model.

## When to use SaaufAI

- [One endpoint for many providers](https://ai.saauf.xyz/docs/api): Use SaaufAI when you want one OpenAI-compatible endpoint (`POST /v1/chat/completions`) for models from OpenAI-compatible hosts, Anthropic and Google, with streaming, tools, images and thinking levels.
- [Per-model data disclosures](https://ai.saauf.xyz/compare): Use it when you must know, before sending data, whether a model trains on it, how long it is kept and where it is processed. Every model has the same YES/NO disclosure sheet.
- [Scoped and rate-limited keys](https://ai.saauf.xyz/docs/api-keys): Use it when each app or agent should get its own key, locked to one model or provider, with a rate limit, spending cap or expiry.
- [Safeguards](https://ai.saauf.xyz/docs/safeguards): Use it when requests should pass rate, token-budget, input-size and content checks before reaching a model, with disclosed fallbacks.

## API

- [OpenAPI spec](https://ai.saauf.xyz/openapi.json): OpenAPI 3.1 description of POST /v1/chat/completions and GET /v1/models, with request, response and error schemas.
- [API reference](https://ai.saauf.xyz/docs/api): Base URL, authentication, Chat Completions, streaming, tools, images, thinking levels, Responses API.
- [Authentication and API keys](https://ai.saauf.xyz/docs/api-keys): Create, scope, limit and revoke keys.
- [Errors and limits](https://ai.saauf.xyz/docs/errors): Every error type, its HTTP status and what to do.
- [Quick start](https://ai.saauf.xyz/docs/quick-start): From signing in to a first API call.

## Models

- [Model directory](https://ai.saauf.xyz/models): Every model with its id, pricing, capabilities and data disclosures.
- [Compare disclosures](https://ai.saauf.xyz/compare): The YES/NO disclosure answers side by side.
- [Companies](https://ai.saauf.xyz/companies): Who makes the models.
- [How disclosures work](https://ai.saauf.xyz/docs/disclosures): What each question means.

## Docs

- [Docs index](https://ai.saauf.xyz/docs): All guides.
- [Introduction](https://ai.saauf.xyz/docs/introduction): What SaaufAI is.
- [Wisps](https://ai.saauf.xyz/wisps): Always-on AI assistants.
- [Homepage as Markdown](https://ai.saauf.xyz/index.md): This site's front page in Markdown.

## Policies

- [About SaaufAI](https://ai.saauf.xyz/about): Who runs it and how.
- [Terms of Service](https://ai.saauf.xyz/policies/terms)
- [Privacy Policy](https://ai.saauf.xyz/policies/privacy)
- [Acceptable Use Policy](https://ai.saauf.xyz/policies/acceptable-use)
- [Security Policy](https://ai.saauf.xyz/policies/security)
- [Model disclosures policy](https://ai.saauf.xyz/policies/model-disclosures)
- [Support and contact](https://ai.saauf.xyz/support): Email ai@saauf.xyz.
- [Security disclosure](https://ai.saauf.xyz/security): How to report a security problem, scope and safe harbour.

## Optional

- [Sitemap](https://ai.saauf.xyz/sitemap.xml)
- [security.txt](https://ai.saauf.xyz/.well-known/security.txt)
- [Blog](https://ai.saauf.xyz/blog)