# walter-grace/mac-code

mac code — Claude Code, but it runs on your Mac for free.   35B AI agent at 30 tok/s via Apple Silicon flash-paging. $0/month.

Repository: https://github.com/walter-grace/mac-code
Canonical: https://ross.abutalabs.com/products/mac-code
Language: Python
License Family: other
Last push: 2026-04-09T18:39:22+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 76, release rhythm 35, longevity 11
- inputs: {"age_days": 163, "days_push": 146, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1026, forks 109 (observed 2026-08-28T04:03:16.777415+00:00)

## What it is
A free, local coding agent for Apple Silicon Macs that runs large quantized LLMs (e.g., Qwen 35B) via llama.cpp, using techniques like flash-paging and SSD weight streaming to run models that exceed RAM. It provides a Claude Code-like terminal agent experience with no subscription cost.

## Use cases
- run a local coding agent on my mac for free
- run a 35B model on a 16GB mac
- replace claude code with a local model
- run llms that don't fit in ram on apple silicon
- stream model weights from ssd to run big models
- self-host an ai coding assistant offline
- get fast local llm inference on mac mini

## When to choose
- you have an Apple Silicon Mac and want a free, private, offline coding agent
- you want to run models larger than your RAM without heavy quantization
- you're comfortable with terminal tools, llama-server, and downloading GGUF models

## When to avoid
- you need top-tier coding quality comparable to frontier cloud models
- you need fast inference for dense models far exceeding RAM (speeds can drop below 1 tok/s)
- you need a permissive license or polished GUI - the repo has no license and is terminal-based

## Facets
- artifact type: application
- maturity: active
- function: llm-inference, agent-framework, cli, chat-interface
- domain: large-language-models, artificial-intelligence, developer-tools
- platform: python, cli
- tags: local-llm, apple-silicon, llama-cpp, gguf, quantization, flash-streaming, coding-agent, free, off-device-ai, moe, ai-agents, command-line, macos, desktop

## Member repositories
- walter-grace/mac-code (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:16.777415+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:08:00.742042+00:00, confidence not recorded.
  - readme: https://github.com/walter-grace/mac-code (fetched 2026-08-28T04:03:16.777415+00:00, sha e3dccbf635df)
- Data as of 2026-08-30T08:39:29.467469+00:00.