Hugging Face MCP ServerUnclaimed

Hugging Face MCP Server connects an LLM to the Hugging Face Hub and to thousands of Gradio AI applications. It is an MCP server, so any client that supports the Model Context Protocol can call its tools to reach Hub search, repository details and other Hub API endpoints from inside a chat or coding session.

The project is aimed at developers and AI tool users who already work with models, datasets and Spaces on Hugging Face and want their assistant to query them directly. The repository describes it as the official Hugging Face server.

Setup is covered on the repository page. Hosted use is possible: the page gives commands and configuration snippets for adding the server to Claude Desktop, Claude Code, Gemini CLI, VS Code and Cursor, with a Hugging Face token used for authentication. For local use, it can run through npx or Docker, in either STDIO mode or stateless Streamable HTTP mode. The HTTP transport starts on port 3000 by default, with a management dashboard at /metrics.

The set of available tools can be changed. A bouquet query parameter exposes additional Hub filesystem, search, Space and Jobs tools, and some of those require authentication. Individual tools can be hidden with the DISABLE_TOOLS environment variable.

The code is released under the MIT license and lives in the huggingface/hf-mcp-server repository on GitHub. The page does not state a price for the server itself, so no cost is listed here.