Hugging Face · docker — Claude Code MCP server: models, datasets, papers
An MCP server that hands your coding agent Hugging Face models, datasets and research papers without leaving the chat. Useful when model code and model data live in the same task.
no license updated 2026-10-29
Open on GitHubWhat Hugging Face does
The server plugs Hugging Face into your agent. Ask about a model card, a dataset or a paper and the MCP client passes the call through, so you stay in the editor.
It runs with Claude, Codex, Cursor, GitHub Copilot, Windsurf, Cline, Zed and Gemini CLI. Docker publishes it. The tools cover models, datasets and research papers.
On a project that never touches Hugging Face the tools just sit idle. Wire the server up where models or datasets actually matter.
What it helps with
- pull a model card into the chat mid-task
- see what a dataset holds without opening a browser
- talk a research paper through with the agent
- keep model and dataset lookups next to the code session
Example prompts for the agent
- Open the Hugging Face model
<model_id>and tell me what task it fits - Check the description of dataset
<dataset_id>and say whether it works for classification - Find a Hugging Face paper about
<topic>and summarise its findings
How to install Hugging Face
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Hugging Face
What does the Hugging Face MCP server do?
It gives the agent tools that reach Hugging Face models, datasets and research papers from inside the chat, so you skip the browser tab.
Which clients can run it?
Claude, Codex, Cursor, GitHub Copilot, Windsurf, Cline, Zed and Gemini CLI.
Who publishes the server?
It comes from docker. You attach it as an MCP server to any of the supported clients.
Similar MCP servers
All MCP servers for AI and ML →Hugging Face by other authors comes first.