Claude Code Local

Claude Code Local lets developers use Claude Code with a model running on their own computer. It suits people who want to keep coding offline or continue working when they hit cloud usage limits.

• Coding tasks: The local model can read and edit files, run commands and work on projects through Claude Code.
• Models that fit your hardware: Automatic model selection matches the Mac's memory to Qwen 3.5, Gemma 4 or Qwen 3.8, with smaller Qwen models available for 8 GB and 16 GB Macs.
• Image input: The small-Mac Qwen models can interpret images as well as handle text and code.
• Optional companion tools: Separate projects add local speech input and spoken replies through NarrateClaude, session management through Moonstone, and browser control through browser-agent.

The project is free and open source under the MIT license. It requires Claude Code and uses MLX on Apple Silicon Macs, or llama.cpp with GGUF models on Linux and Windows. Model files come from Hugging Face; once downloaded, local inference works offline without a company account or API key. Code stays on your machine, and the project reports no telemetry.

Memory affects which model you can run. On 8 GB and 16 GB Macs, the smaller models have limited context and are intended for requests with one clear step at a time.