For developers

OpenAI-compatible.
Agent-capable.
Mesh-routed.

A drop-in OpenAI / Ollama / Anthropic server on 127.0.0.1:1473, a tool-calling chat agent, and a private mesh that routes inference between your devices. No broker. No account. No API key.

Drop-in for any OpenAI client

127.0.0.1:1473 speaks OpenAI, Ollama, and Anthropic. The chat agent ships with tools for your mesh, filesystem, and shell, all behind per-device permission gates, persisted across reloads. Tools execute on the caller's box even when inference is on a remote peer, so your Pi can borrow the workstation's model and still configure the Pi. Cursor, Continue, Aider, Cline, Zed, Open WebUI, opencode all just work.

Developer docs → GitHub 

# point any OpenAI client at your machine
curl http://127.0.0.1:1473/v1/chat/completions \
  -H "Authorization: Bearer myownllm" \
  -d '{
    "model": "myownllm",
    "messages": [{"role":"user","content":"hi"}]
  }'

Local AI. Voice Mode. Private Mesh.

Because talking privately to your own AI shouldn't be hard. Free, open source, one click to install.

Download MyOwnLLM