概览
What is HuggingMCP?
HuggingMCP is a comprehensive Model Context Protocol (MCP) server that allows Claude and other MCP-compatible AI assistants to interact seamlessly with the Hugging Face ecosystem. It enables natural language management of models, datasets, files, pull requests, and collections.
How to use HuggingMCP?
Install the server in a Python 3.10+ environment with uv or pip, then configure your Hugging Face token and permission settings in the Claude Desktop configuration file. After restarting Claude, you can use natural language commands to perform operations like creating repositories, editing files, searching models, and managing pull requests.
Key features of HuggingMCP
- Create, delete, and manage Hugging Face repositories
- Create and manage pull requests with file uploads
- Read, write, edit, and delete files on repositories
- Search models, datasets, and spaces with filters
- Organize repositories into collections
Use cases of HuggingMCP
- Use Claude to create a new model repository with a custom README
- Edit a training script by replacing a specific line of code
- Search for the top 10 most downloaded text classification models
- Create a pull request to improve a model card on a public repo
- Organize favorite models and datasets into a personal collection
FAQ from HuggingMCP
What permissions are available?
Read-only, write-only, and admin mode are set via environment variables (HF_READ_ONLY, HF_WRITE_ONLY, HF_ADMIN_MODE). Default mode provides full read/write access.
What Python version is required?
Python 3.10 or higher is required.
How do I enable admin mode for deleting repositories?
Set the HF_ADMIN_MODE environment variable to "true" in the Claude Desktop configuration.
How can I test that the server is connected properly?
Look for the 🔨 hammer icon in Claude Desktop or use the hf_test tool to verify the connection.
Where can I find debug logs if something goes wrong?
On macOS: ~/Library/Logs/Claude/mcp.log; on Windows: %APPDATA%/Claude/Logs/mcp.log. You can also enable verbose logging in main.py.