Huggingface MCP Server

🚀 🤗 Hugging Face MCP Server 🤗

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

📋 Overview

🧩 Components

📦 Resources

• Custom hf:// URI Scheme: Used to access various resources on Hugging Face.
• Models: Include models of various sizes and types, such as BERT, GPT, etc.
• Datasets: Provide datasets for various tasks like text classification and image generation.
• Spaces: Showcase user-created applications using Gradio or other frameworks.
• Papers: Access the latest AI research papers, filtered by date or topic.

💡 Prompt Templates

• compare-models: Compare the performance and characteristics of two or more models. For example, compare the Llama-3-8B and Mistral-7B models.
• summarize-paper: Summarize the content of a paper with a specified arXiv ID. For example, use arXiv ID 2307.09288 to get the abstract.

🛠️ Tool Categories

• Model Tools: Such as search-models, get-model-info, etc.
• Dataset Tools: Such as find-datasets, popular-text-classification, etc.
• Space Tools: Such as show-popular-spaces, image-generation, etc.
• Paper Tools: Such as today-featured-papers, summarize-paper, etc.
• Collection Tools: Such as find-collections, created-by-user, etc.

⚙️ Configuration

• No Configuration Required: Use the server's features directly.
• Optional Hugging Face API Token: Increase API request rate and permissions. Set the environment variable HF_TOKEN.

🚀 Quick Start Guide

📦 Installation

Install via Smithery:

# Example command

Configuration file locations for different operating systems:

• macOS: ~/Library/Logs/Claude/mcp-server-huggingface.log
• Windows: %APPDATA%\Claude\logs\mcp-server-huggingface.log

👨‍💻 Development

🛠️ Build and Publish

Use the uv tool:

1. Synchronize dependencies:

uv sync

2. Build the package:

uv build

3. Publish:

uv publish

🐞 Debugging

Use the MCP checker:

# Example command

Access the provided URL to start debugging.

💡 Example Prompts

• Claude Examples:
  • "Search for BERT models on Hugging Face with less than 100 million parameters."
  • "Find the most popular datasets for text classification tasks."

🚧 Troubleshooting

1. Check Logs: View the log files for the corresponding platform.
2. Handle Rate Limits: Add an HF_TOKEN to increase permissions.
3. Network Issues: Ensure the device is connected to the internet.
4. Tool Malfunctions: Verify that the data exists on the Hugging Face website.

With the above content, you can fully understand and use the functions of the Hugging Face MCP Server.