mlx-serve
ddalcu/mlx-serve
mlx-serve is a native Zig inference server that runs LLMs on Apple Silicon, supporting both MLX-format models and GGUF files from Hugging Face. It exposes OpenAI-, Anthropic-, and Ollama-compatible HTTP APIs on localhost, so existing clients like Claude Code, the OpenAI SDK, Cursor, and Open WebUI work without changes. It also ships MLX Core, a signed macOS menu-bar app with chat, agent mode, and MCP tool calling. Built for Apple Silicon developers who want local inference without Python or the cloud.
安装方式
brew install --cask mlx-core项目说明
OpenAI- and Anthropic-compatible local inference for Apple Silicon — MLX and GGUF — faster than LM Studio on identical MLX weights. No Python. No cloud. No Electron.