Shreyaskarnik Huggingface MCP Server

🚀 🤗 Hugging Face MCP Server 🤗

[This is a Model Context Protocol (MCP) server that provides read-only access to the Hugging Face Hub API. It enables large language models (LLMs) like Claude to interact with Hugging Face's models, datasets, spaces, papers, and collections.]

🚀 Quick Start

📦 Installation

1. Via Smithery CLI:

   smithery install huggingface-mcp-server
   

2. Via Claude Desktop:
   • macOS: Add the server to ~/Library/Application Support/Claude/
   • Windows: Add the server to %APPDATA%\Claude\

💻 Usage Examples

Here are some example prompts for Claude:

• "Search for BERT models on Hugging Face with less than 100M parameters."
• "Find the most popular text classification datasets on Hugging Face."
• "What are the featured AI research papers on Hugging Face today?"
• "Summarize the paper with arXiv ID 2307.09288."
• "Compare the Llama-3-8B and Mistral-7B models on Hugging Face."
• "Show the most popular Gradio spaces for image generation on Hugging Face."
• "Find collections created by TheBloke that contain Mistral models."

✨ Features

Resources

• Hugging Face Models: Provide access to Hugging Face models, e.g., via the hf://model/{model_id} URI scheme.
• Datasets: Provide access to Hugging Face datasets, e.g., via the hf://dataset/{dataset_id} URI scheme.
• Spaces: Provide access to Hugging Face spaces (such as Gradio and Streamlit apps).
• Papers: Provide access to AI papers and research hosted on Hugging Face.
• Collections: Provide access to collections of models and datasets created on Hugging Face.

Prompt Templates

• compare-models: Compare two or more models (e.g., compare Llama-3-8B and Mistral-7B).
• summarize-paper: Summarize a paper with a specific arXiv ID (e.g., arXiv ID 2307.09288).

Tools

• Model Tools:
  • search-models: Search for models on Hugging Face (e.g., search for BERT models with less than 100M parameters).
  • get-model-info: Get detailed information about a specific model (e.g., get information about Llama-3-8B).
• Dataset Tools:
  • search-datasets: Search for datasets on Hugging Face (e.g., search for datasets for text classification tasks).
  • get_dataset_info: Get detailed information about a specific dataset.
• Space Tools:
  • search-spaces: Search for Hugging Face spaces (e.g., search for Gradio spaces in the field of image generation).
  • get-space-info: Get detailed information about a specific space.
• Paper Tools:
  • search-papers: Search for AI papers on Hugging Face (e.g., search for featured research papers of the day).
  • get-paper-info: Get detailed information about a specific paper.
• Collection Tools:
  • search-collections: Search for collections on Hugging Face (e.g., search for collections created by TheBloke that contain Mistral models).
  • get-collection-info: Get detailed information about a specific collection.

📚 Documentation

Configuration

This server does not require configuration, but it supports optional Hugging Face API authentication:

• Set the environment variable HF_TOKEN to increase API request rate limits and access restricted content.

Troubleshooting

1. Check Server Logs:
   • macOS: ~/Library/Logs/Claude/mcp-server-huggingface.log
   • Windows: %APPDATA%\Claude\logs\mcp-server-huggingface.log
2. Handle API Rate Limit Errors: Consider setting a Hugging Face API token.
3. Ensure a Stable Network Connection.
4. Problems with Specific Tools: Check if the same data exists on the Hugging Face website.

Development

1. Build the Server:

   uv sync && uv build
   

2. Publish the Server:

   uv deploy
   

3. Debugging:
   • Use the MCP Inspector for debugging:

     inspector run --server huggingface-mcp-server
     

📄 License

MIT License

👥 Contributors

• [Your Name]