scouzi1966Community·10 Agent Skills Included

maclocal-api

Local OpenAI-compatible model server for Apple Silicon Macs

Runs local language models on Apple Silicon through an OpenAI-compatible API with streaming, tool calling, structured output, and reasoning extraction. Eliminates cloud API costs, API keys, and data privacy concerns by serving MLX and Apple Foundation Models directly from your Mac. Includes skills to build, test, benchmark, add new models, and publish releases faster with automated validation.

324|17

↗

npx skills add scouzi1966/maclocal-api --all -g -y

Available:

Gives your AI agent the build commands, architecture map, and runtime rules needed to compile, run, and test this local model server without breaking its strict dependency boundaries.

All Skills in This Repository (10)

Pure Emerald Level Indicators

📦 In Repo

scouzi1966

test-macafm

Run maclocal-api test suites with pass/fail assertions and AI-judge analyses.

build-afm-nightly-publish

Automate the AFM nightly release pipeline from build to Homebrew tap updates.

afm-build-promote-nightly

Build AFM stable releases from main HEAD or nightly commits with validated patches and packaging.

afm-release-wheel

Build and publish a PyPI wheel from an existing AFM binary.

test-opencode-tooling

Classify tool-call failures between OpenCode and afm during automated testing.

test-afm-binary

Validate pre-built afm binaries with pre-flight checks and test suites on macOS.

build-afm

Build the AFM binary from a clean clone with submodules, patches, web UI, and Swift.

add-afm-model

Analyze HuggingFace MLX model config.json files to onboard them into AFM registries and Swift backend workflows.

codex-promptfoo-agentic-eval

Run and interpret Promptfoo-based AFM agentic evaluation suites for QA.

afm

Launch a local macOS ML stack with an OpenAI-compatible API.

Community

Advanced

Frequently Asked Questions

FAQPage Schema

How to install maclocal-api?▼

Run `npx skills add scouzi1966/maclocal-api --all -g -y` in your terminal to install all skills in this suite globally.

How to run local models on a Mac with an OpenAI API?▼

Install the afm binary via Homebrew or pip, then run `afm mlx -m <model-id>` to serve any Hugging Face MLX model at a local OpenAI-compatible endpoint.

Does maclocal-api support tool calling and streaming?▼

Yes. It supports OpenAI-compatible tool calling with multiple parser formats, streaming SSE responses, structured JSON output, logprobs, and reasoning extraction.

Can I add a new Hugging Face model to AFM?▼

Yes. The add-afm-model skill checks model compatibility, registers new architectures, and ports implementations so your agent can onboard new MLX models.

Does maclocal-api work with coding agents like OpenCode?▼

Yes. It is built for tool-using clients and includes dedicated test skills that validate tool call reliability with OpenCode, OpenClaw, and other agent frameworks.