LiteLLM proxy with PicoClaw
LiteLLM exposes an OpenAI-shaped API across dozens of backends. PicoClaw can treat it like any other OpenAI-compatible endpoint—useful when you want fallbacks, spend tracking, or model aliases in one place.
1. When LiteLLM helps
- Multiple vendor keys behind one gateway
- Routing rules and budgets per model
- Staging parity with production model names
2. Deployment
Run LiteLLM in Docker on a small VM or next to PicoClaw on your LAN. Lock down network access; the proxy holds powerful credentials.
3. Point PicoClaw at the proxy
Set api_base to your LiteLLM URL and choose model aliases you defined. Confirm with a single picoclaw agent smoke test before enabling gateways.