# Omni Octo API docs

> Developer documentation for the Omni Octo API: the OpenAI-compatible /v1 API, the MCP servers, and the platform REST API.

> Machine-readable: full text at https://docs.omniocto.com/llms-full.txt · any page as markdown by appending .md · OpenAPI at https://docs.omniocto.com/openapi/v1.json and https://docs.omniocto.com/openapi/platform.json

## Getting started
- [Get an API key](https://docs.omniocto.com/getting-started/api-keys/): Create, reveal, and revoke workspace API keys from the dashboard Settings page.
- [Authentication](https://docs.omniocto.com/getting-started/authentication/): How API keys map to your workspace, role, and permissions across every surface.
- [Overview](https://docs.omniocto.com/getting-started/overview/): What the Omni Octo API is, the three ways to integrate, and where to go next.

## OpenAI-compatible API
- [Confirmations](https://docs.omniocto.com/openai-api/confirmations/): How risky tools park for confirmation over the API, and why stateless calls can't confirm them.
- [Errors](https://docs.omniocto.com/openai-api/errors/): The OpenAI error envelope and the full status/type/code/SDK-exception table.
- [Events & webhooks](https://docs.omniocto.com/openai-api/events/): Subscribe to workspace state changes via outbound webhooks or the GET /v1/events SSE feed — including changes your agent did not initiate.
- [OpenAI-compatible API overview](https://docs.omniocto.com/openai-api/overview/): The /v1 facade — endpoints, base URL, and the curated model allowlist.
- [Rate limits](https://docs.omniocto.com/openai-api/rate-limits/): Per-key and per-workspace request budgets, and how 429s are signaled.
- [Statefulness](https://docs.omniocto.com/openai-api/statefulness/): previous_response_id, the conversation extension param, and stateless calls compared.
- [Streaming](https://docs.omniocto.com/openai-api/streaming/): Server-Sent Events for /v1/responses and /v1/chat/completions, and the server-side settle guarantee.
- [Usage & billing](https://docs.omniocto.com/openai-api/usage/): How token usage is mapped per endpoint, and what happens when a workspace is suspended.

## MCP servers
- [Connect an MCP client](https://docs.omniocto.com/mcp/connect/): Point a generic MCP client at Octo's public or meta-agent MCP server with a workspace API key.
- [MCP servers overview](https://docs.omniocto.com/mcp/overview/): The two MCP servers Octo exposes — the public tool catalog and the meta-agent conversation bridge.
- [Tool catalog](https://docs.omniocto.com/mcp/tools/): Every tool Octo exposes over MCP — 206 in total, with the risk class that decides whether it runs immediately or waits for a human.

## Platform REST API
- [Platform REST API overview](https://docs.omniocto.com/platform-api/overview/): The broader /api/* dashboard API for advanced, direct resource access.

## For agents
- [Agent resources](https://docs.omniocto.com/agents/resources/): Machine-readable artifacts for AI agents consuming these docs — llms.txt, per-page markdown, OpenAPI JSON, and a live docs-search MCP.

## Changelog
- [Changelog](https://docs.omniocto.com/changelog/): Notable changes to the Omni Octo API, MCP servers, and developer docs.

## Reference
- [OpenAI-compatible API (/v1) — interactive](https://docs.omniocto.com/openapi/v1/)
- [Platform REST API — interactive](https://docs.omniocto.com/openapi/platform/)
- [Docs-search MCP](https://api.omniocto.com/mcp-docs) — query these docs live from an agent