Claude Code Local
A small local server that runs open-weight AI models on an Apple Silicon Mac's own chip and answers in Anthropic's API format, so Claude Code can work with no cloud model and no API key.
Claude Code Local is a small Python server that runs open AI models on a Mac's own Apple Silicon chip through Apple's MLX framework, and answers Claude Code in Anthropic's own Messages API format with streaming, so Claude Code works without a cloud model or an API key. It supports several open-weight models, including Muse-Glimmer 30B, Gemma 4 31B, Qwen 3.5 122B and DeepSeek V4 Flash, translating each model's own tool-call format into Claude Code's tool calls, repairing garbled tool-call JSON, and reusing the prompt cache between turns so long sessions do not re-read everything.
The repository also includes a one-file terminal coding agent with a short, fixed system prompt designed to keep its cache warm across turns, tests for the tool-call parser and the running server, install and setup scripts that pick a model to fit the Mac's memory, and desktop launcher scripts.
It is written in Python and released under the MIT license.