mlintern

mlintern Claude Plugin

Use ml-intern from inside Claude Code.

ML Intern is an open-source ML engineer agent from Hugging Face that can research, write, and run ML workflows such as fine-tuning, evaluation, and shipping models. This plugin makes it easy to invoke that workflow from Claude Code.

Experimental

This plugin is experimental and may change frequently. Expect rough edges.

What this plugin provides

• /mlintern "fine-tune a model" to run an ML Intern task
• /mlintern --background "..." to run in background
• /mlintern:status to check running/recent jobs
• /mlintern:result [job-id] to fetch final output
• /mlintern:cancel [job-id] to cancel a running job
• /mlintern:setup to verify local setup

Requirements

• Node.js 18+ (if you are already using Claude Code, you should already have this)
• ml-intern installed and available on PATH
• Auth configured for ml-intern (HF token and any model provider keys you use)

The plugin install step itself does not currently install ml-intern for you. That keeps plugin install predictable across environments, but means ml-intern still needs to be installed separately on your machine.

Install in Claude Code

The /plugin ... .

Install ml-intern first (skip if already installed):

git clone git@github.com:huggingface/ml-intern.git
cd ml-intern
uv sync
uv tool install -e .

Then install this Claude plugin (the commands below are Claude Code slash commands and should be run inside an active Claude Code session, not your regular shell terminal):

/plugin marketplace add huggingface/mlintern-plugin
/plugin install mlintern@huggingface-mlintern
/reload-plugins
/mlintern:setup

Run it!

Now, in your Claude Code session, you can do:

/mlintern --background "Find an Urdu speech dataset and fine-tune a small Whisper model in under 1 hour, evaluate WER, and share the model link and metrics.
/mlintern:status
/mlintern:result