This project provides a server to connect Claude Desktop with Hugging Face Spaces, enabling easy integration of various AI models for tasks like image generation, vision, and text-to-speech. It targets users of Claude Desktop seeking to leverage a wide array of Hugging Face models with minimal configuration.

How It Works

The server acts as a bridge, translating Claude's Model Context Protocol (MCP) requests into calls to Hugging Face Spaces. It automatically discovers and configures endpoints for specified Spaces, handling file uploads/downloads and model outputs. The default "Claude Desktop Mode" integrates model outputs directly into Claude's context or saves files to a specified working directory, optimizing the user experience.

Quick Start & Requirements

• Install via NPM: @llmindset/mcp-hfspace
• Requires NodeJS and Claude Desktop 0.78 or greater.
• Configuration involves adding the server to claude_desktop_config.json.
• Official documentation: llmindset.co.uk/resources/mcp-hfspace/

Highlighted Details

• Supports image generation, vision models, text-to-speech, speech-to-text, and chat models.
• Handles file I/O via a configurable working directory.
• Supports private Hugging Face Spaces with a provided token.
• Can specify exact API endpoints within a Space.

Maintenance & Community

• Developed by evalstate.
• No explicit community links (Discord/Slack) or roadmap are provided in the README.

Licensing & Compatibility

• The README does not explicitly state a license. The NPM package name suggests it's available via NPM.

Limitations & Caveats

• Endpoints with unnamed parameters are unsupported.
• Complex Python type translations to MCP formats may have limitations.
• Claude Desktop's error handling and timeouts (around 60s) can affect long-running jobs on ZeroGPU Spaces.
• Hugging Face ZeroGPU quotas may be a bottleneck; using a personal HF token for dedicated hardware is recommended for production.