Run Claude Code 100% on-device: MLX-native server for Apple Silicon hits 3.3k GitHub stars
tom_doerr · x · 2026-09-10
claude-code-local (3.3k stars on GitHub) is an MLX-native server that speaks the Anthropic API, letting you run Claude Code entirely on Apple Silicon — no cloud, no API key, airgap-ready for NDA/legal/healthcare workflows. Ships with swappable models including Hermes 4 14B, Gemma 4 31B, multimodal abliterated Muse-Glimmer 30B, Llama 3.3 70B, Qwen 3.5 122B (65 tok/s), and DeepSeek V4 Flash with 1M context, plus leaderboards, voice, browser, phone and MCP support.
More from coding & agent
- OpenAI demos Astra driving Blender to build editable 3D scenes from prompts — xiaohu · 2026-09-10
- OpenAI case study: GPT-6 Astra builds a house in Blender from one prompt, then moves it into UE5 — xiaohu · 2026-09-10
- Astra Prompt Studio: open-source prompt generator turns goals into structured agent prompts — Scobleizer · 2026-09-10
- Sentry founder on eval costs: keep asserts deterministic, LLM judges get brutal — zeeg · 2026-09-10
- Dev builds local 'librarian' plugin to stop AI agents trashing repos like raccoons — Deep_Bus_7488 · 2026-09-10
- Building an internal multi-agent system keeps breaking; V3 already planned — canuck_amok · 2026-09-10