Get started

Install uv

MedMLX needs macOS 14.0 or later on Apple Silicon. It uses uv to install Python 3.13 and its packages. If uv --version already works, skip this step.

curl -LsSf https://astral.sh/uv/install.sh | sh

With Homebrew, use brew install uv instead. Open a new terminal window afterwards.

Install MedMLX

This installs the medmlx command, its MCP server and every model family.

uv tool install --python 3.13 "medmlx[all,mcp]"

For a smaller install, use segment, generate or vision instead of all. For a Python project, use uv add "medmlx[all,mcp]". If your shell cannot find medmlx, run uv tool update-shell and open a new terminal window.

Run a model

describe lists the arguments a model takes. Save yours in arguments.json. readiness checks that the model can run, without downloading anything. run downloads the weights the first time, then runs the model.

medmlx describe mlx-segment-ct
medmlx readiness mlx-segment-ct --arguments arguments.json --root /path/to/work
medmlx run mlx-segment-ct --arguments arguments.json --root /path/to/work

To download weights ahead of time for offline use:

medmlx setup --all --progress

MedGemma runs in a separate med-vlm-mlx environment and needs local weights. MRSegmentator also needs local weights. Pass the folder as weights_path.

Command linePython API

Connect an AI agent (optional)

An MCP client can list, check and run models for you. Give this command to your client as the server to start. The folder must already exist:

medmlx serve --root /path/to/work

MCP guide