Claude Code Local is a local AI server that runs Claude Code entirely on-device using Apple Silicon hardware with MLX-native implementation of the Anthropic API.
The tool addresses the need for private, offline code execution in environments where data cannot leave the device. It works by running multiple large language models locally on Apple Silicon Macs, eliminating dependency on cloud services. The project includes six model options spanning different capability and speed tradeoffs, including Muse-Glimmer 30B with multimodal image reading capabilities, Gemma 4 31B, Qwen 3.5 122B optimized for throughput, and DeepSeek V4 Flash with extended context windows. By implementing an MLX-native Anthropic API server, it allows existing Claude Code workflows to function entirely offline.
This tool suits organizations handling sensitive data under regulatory constraints such as healthcare, legal, or NDA-bound work where data residency and airgap requirements are non-negotiable. It is purpose-built for developers on Apple Silicon who need code generation capabilities without cloud transmission. The project explicitly targets private and offline-first use cases rather than serving as a general-purpose local LLM runner.
The project shows active development with recent additions of multimodal capabilities to its flagship model and ongoing model integration work. Maintenance activity includes regular updates to supported model options and performance optimizations for Apple Silicon execution.
