mlx-vlm
MLX-VLM is a package for inference and fine-tuning of Vision Language Models (VLMs) on your Mac using MLX.
MLX-VLM agent skills provide Claude Code, Codex, and Gemini Code agents with structured knowledge for running inference, converting models, adding architectures, benchmarking, and debugging MLX-VLM Vision Language Models on Mac. The skills bundle includes seven specialized workflows covering CLI inference, server deployment, model quantization, architecture porting, performance testing, contribution guidelines, and issue reporting. By loading these skills, coding agents follow correct MLX-VLM project conventions instead of guessing command syntax or configuration patterns.
Key Features
Use Cases
- 01Enabling Claude Code to correctly run MLX-VLM inference commands with proper flags for text, image, and audio inputs
- 02Guiding agents through Hugging Face model conversion to MLX format with correct quantization parameters
- 03Helping coding agents port new vision-language architectures into the MLX-VLM framework
- 04Assisting with reproducible performance benchmarking and fork-vs-main comparisons for pull requests
- 05Teaching agents MLX-VLM contribution standards including code placement, pre-commit hooks, and PR expectations
- 06Structuring debugging workflows to generate actionable GitHub issues from CLI or server failures
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
mlx-vlm — FAQ
What are MLX-VLM agent skills?+
A bundle of seven specialized skill definitions that teach Claude Code, Codex, and Gemini coding agents how to correctly use MLX-VLM for vision-language model inference, conversion, development, and support tasks on Mac.
How do I install the MLX-VLM skills in Claude Code?+
Clone the mlx-vlm repository locally, then run '/plugin marketplace add /path/to/mlx-vlm' followed by '/plugin install mlx-vlm-skills@mlx-vlm' in Claude Code.
Which coding agents support these skills?+
The skills work with Claude Code (via /plugin commands), Codex CLI (codex plugin add), and Gemini CLI (gemini extensions install).
Do I need API keys or special prerequisites?+
You need a local clone of the mlx-vlm repository and the corresponding coding agent (Claude Code, Codex, or Gemini) already installed. No additional API keys are required for the skills themselves.
Are the MLX-VLM agent skills free?+
Yes, the skills bundle is open source and ships with the mlx-vlm repository under skills/ at no cost.
Can I validate the skills before installing?+
Yes, run 'python3 skills/scripts/validate_skills.py' from the repository root to check skill integrity before installation.
How do I install mlx-vlm?+
Open the source repository on GitHub and follow its README. mlx-vlm is a skill — MCP Agents Market links you directly to the official repo.
Is mlx-vlm free?+
mlx-vlm is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.