Hugging Face

This is Hugging Face’s official MCP Server that connects your AI assistants directly to the Hugging Face Hub ecosystem.

It enables you to search, explore, and interact with thousands of machine learning models, datasets, and research papers without leaving your MCP clients.

Features

• 🔍 Semantic Search for Spaces – Find AI applications using natural language queries
• 📚 Research Paper Discovery – Search ML research papers with intelligent matching
• 🤖 Advanced Model Search – Filter models by task, library, framework, and performance metrics
• 📊 Detailed Model Information – Access comprehensive model specifications, usage stats, and documentation
• 🗃️ Dataset Exploration – Search datasets with author, tag, and category filters
• 📋 Dataset Details – Retrieve complete dataset information including structure and licensing
• 🔐 Secure Authentication – Token-based access using Hugging Face credentials
• ⚡ Real-time Integration – Direct API connection without manual downloads or configurations

Use Cases

• Model Selection for Production: Quickly compare and evaluate models for specific tasks like text classification or image recognition, accessing performance metrics and community ratings to make informed decisions
• Research and Development: Discover cutting-edge papers and corresponding model implementations, enabling rapid prototyping and experimentation with latest ML techniques
• Dataset Curation: Find and analyze datasets that match your project requirements, including checking licensing terms and data quality metrics before integration
• AI Application Discovery: Explore Spaces to find working examples and demos that demonstrate specific AI capabilities, helping you understand implementation patterns and user experience approaches

How to Use It

1. You will need to generate a “Read” access token from your Hugging Face account settings. This token is necessary for the server to authenticate your requests.

2. Configure Your MCP Client:

VS Code:

{
  "servers": {
    "hf-mcp-server": {
      "url": "https://huggingface.co/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_HF_TOKEN"
      }
    }
  }
}

Cursor:

{
  "mcpServers": {
    "hf-mcp-server": {
      "url": "https://huggingface.co/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_HF_TOKEN"
      }
    }
  }
}

Claude Desktop:

{
  "mcpServers": {
    "hf-mcp-server": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://huggingface.co/mcp",
        "--header",
        "Authorization: Bearer YOUR_HF_TOKEN"
      ]
    }
  }
}

Other Clients:

For other MCP-compatible clients, you will need to provide your Hugging Face Read token to establish the connection similarly.

Built-in Tools

The server comes with several pre-configured tools to interact with the Hugging Face Hub:

• Spaces Semantic Search: Find AI applications using natural language queries.
• Papers Semantic Search: Search for machine learning research papers.
• Model Search: Filter and search for models based on tasks, libraries, and other criteria.
• Model Details: Get comprehensive information about a specific model.
• Dataset Search: Find datasets using filters for authors, tags, and more.
• Dataset Details: Retrieve detailed information about a particular dataset.

FAQs

Q: Can I access private models and datasets?
A: Yes, if you use a valid Hugging Face API token with the necessary permissions, you can access your private resources.

Q: What if I’m having trouble connecting?
A: First, double-check that your API token is correct and has “Read” permissions. Also, ensure the JSON configuration in your client is correctly formatted and that you have restarted your client after making changes.

Q: How does the semantic search differ from regular keyword search?
A: Semantic search understands the meaning behind your queries, so you can search using natural language like “models for sentiment analysis of social media posts” instead of exact keyword matches.