Skip to content

Felix Documentation

Felix is a self-hostable managed agents harness. You author agents as YAML (apiVersion: felix/v1); Felix compiles them into runnable agents with governance, durable execution, and multi-protocol surfaces.

  • OpenAI-compatible chat completions at /v1/chat/completions (sync + SSE)
  • A2A JSON-RPC at /a2a
  • MCP HTTP JSON-RPC at /mcp
  • Direct REST/SSE at /chat

Management surfaces (/audit, /plans, /jobs, /approvals, /manifests, /eval, /usage) cover observability, HITL, canary manifests, and offline eval.

Runtime: felix-run/felix (CPython, Compose / Helm).
Chat UI + this docs site: felix-run/web (Cloudflare Workers frontends).

Commerce is out of scope for this harness.

Three seams

  • Session — append-only events with pluggable strategies (full_replay / windowed / summarizing / semantic / compacting).
  • Pattern / model registries — open registration for loops (react, deep, …) and providers (anthropic, openai, ollama, workers_ai, groq, …), with the model layer in its own package that cannot import the harness.
  • ToolExecutorlocal / mcp / a2a / container / queue / sandbox / browser / client, wrapped by governance.

Live reference

Running deployment

A running Felix API exposes:

  • GET /openapi.json — OpenAPI 3.x
  • GET /docs — interactive API docs
  • GET /health — liveness
  • GET /.well-known/agent-card.json — A2A discovery

Where to start