Free Mac app 'LocalLM Lab' exposes on-device LLMs as an API server
A free app called LocalLM Lab has been released for running on-device AI on Macs. It started as a simple prompt screen for poking at Apple's built-in on-device . Since then, an OpenAI API-compatible endpoint was added, reachable via localhost or an external IP, so existing tools and scripts that expect an OpenAI-style API can be pointed at it like any other .
Right now it runs on Apple's framework, but thanks to the new protocol Apple introduced this year, the app is no longer locked into Apple's own model. Any provider that ships a conforming Swift package — local MLX models, Claude, Gemini, and others — could plug in. The developer is looking into wiring that up so the app isn't limited to the built-in model, and is asking the community what should come next, including MCP support and in-UI model switching.
The app is a free download from thisbrain.ai/locallm.
Key points
- Free app 'LocalLM Lab' for Macs, downloadable at thisbrain.ai/locallm
- Ships an OpenAI API-compatible endpoint accessible via localhost or an external IP
- Currently runs on Apple's () framework
- Apple's new protocol could let it plug in local MLX models, Claude, Gemini, and more later
- Developer is soliciting feedback on next steps like MCP support and in-app model switching