#139139th of 149 in Skills, plugins and rules

claude-code-local

Anthropic-API MLX server that runs Claude Code on local models on Apple Silicon

Stars: 3.4k
License: MIT
Last commit: Oct 2026
Last release: Aug 2026
Language: Python

Missed the latest GitHub check; these facts may be older.

Overview

A Python server (proxy/server.py) runs open models with Apple MLX and answers Claude Code in the Anthropic Messages API format, with streaming. It converts Gemma, Qwen and Llama-style tool calls into Claude Code tool calls, repairs malformed tool JSON, and reuses the prompt cache between turns. setup.sh picks a model based on the Mac's memory and creates a Desktop launcher.

Who it is for: Mac users who want Claude Code on local models, offline or past usage limits

Strengths

Speaks the Anthropic API directly, with no LiteLLM or OpenAI translation layer
setup.sh picks a model by RAM, from Hermes 4 14B at 16 GB to Qwen 3.8 27B at 96 GB
Launchers disable Claude Code's background traffic; README shows an lsof check for network connections
Includes tool-call parser tests and a benchmark script for its cache numbers

Weaknesses

Apple Silicon Macs only (M1 or newer)
Under 16 GB, the README says tool use is unreliable with Gemma 4 E4B
Benchmark scores come from the Agent-12 harness, not from inside Claude Code
Browser, voice and phone modes need separate repos or tools that setup.sh does not install

What it needs

RAM ≥ 16 GB
GPU required
Needs Claude Code (npm), Python 3.12, mlx-lm
Models: Gemma 4, Qwen 3.8, Hermes 4 14B, MLX
port 4000

Also in Skills, plugins and rules

See all 149

Also in Skills, plugins and rules
Rank	Project	Score
1	cavemanCuts token use in coding agents with terse replies and compressed inputs Live demo ↗ (opens in a new tab)112.2k starsApache-2.0	95 out of 100
2	skillsAgent skills for planning, TDD, debugging and code review292.4k starsMIT	94 out of 100
3	eccAgents, skills, hooks and rules bundle for Claude Code, Codex and other harnesses280k starsMIT	94 out of 100
#4	impeccableDesign skill with 24 commands and a detector for AI-generated frontends Live demo ↗ (opens in a new tab)81.5k starsApache-2.0	94 out of 100
#5	ponytailSkill prompt that makes coding agents write less code and add tests164.1k starsMIT	92 out of 100

How to choose

Check which agent it targets and whether it works with the version you run.
Read what it adds to every prompt; big rule sets cost tokens on each call.
Prefer add-ons you can turn off per project.

Health

Releases: Not checked
Top contributor: Not checked
Security advisories: Not checked
Package advisories: Not checked
OpenSSF Scorecard: Not checked

Not read for this project yet: each refresh reads the facts of a few projects. Bots and coding-agent accounts are left out of the commit counts. Not part of the score: how we check.

License in plain words

MIT Permissive

Use it at work: Yes
Change it and rebrand it: Yes
Sell it or host it for others: Yes

Your part: Keep the copyright and license notice in copies you ship.

A plain summary, not legal advice. Read the license on GitHub ↗ (opens in a new tab)

#139 of 149 in Skills, plugins and rules on Archestack, as of 15 Oct 2026.

In your README

The badge updates twice a day.

[![Archestack](https://archestack.com/badge/nicedreamzapp/claude-code-local.svg)](https://archestack.com/best-of-vibe-coding/projects/claude-code-local/)

How badges work

On your site or blog

A card with the rank, score and signals, in light or dark to match the reader.

<iframe src="https://archestack.com/embed/best-of-vibe-coding/claude-code-local/" title="claude-code-local on Archestack" width="400" height="270" style="border:0;max-width:100%" loading="lazy"></iframe>

Preview the card

Best of Vibe Coding on GitHub

Each review comes from the project's README and facts we check on GitHub. The fix form opens with claude-code-local filled in.