# OpenAI & Anthropic compatible API

Point the SDKs you already use at Olaa and get an AI employee instead of a bare model.

Base URL: `https://api.olaa.com/api/compat`

## OpenAI style

Base URL `https://api.olaa.com/api/compat/v1`, model `olaa`. Routes: POST /chat/completions, POST /responses, GET /models. Streaming supported.

```
const client = new OpenAI({ baseURL: 'https://api.olaa.com/api/compat/v1', apiKey: process.env.OLAA_API_KEY });
await client.chat.completions.create({ model: 'olaa', messages: [{ role: 'user', content: 'Reconcile Stripe and HubSpot for August' }] });
```

## Anthropic style

Base URL `https://api.olaa.com/api/compat`. Routes: POST /messages, POST /messages/count_tokens. Use `x-api-key`.

## Semantics

Every call is a full agent run with a 10-minute cap and consumes credits. System messages are appended to Olaa’s own instructions. `previous_response_id` continues the same thread. Files come back as links in the final message.