Hugging Face

Hugging Face Official MCP Server

Welcome to the official Hugging Face MCP Server 🤗. Connect your LLM to the Hugging Face Hub and thousands of Gradio AI Applications.

Installing the MCP Server

Follow the instructions below to get started:

Install in Claude Desktop or claude.ai

Click here to add the Hugging Face connector to your account.

Alternatively, navigate to https://claude.ai/settings/connectors, and add "Hugging Face" from the gallery.

claude mcp add hf-mcp-server -t http https://huggingface.co/mcp?login

claude mcp add hf-mcp-server \
  -t http https://huggingface.co/mcp \
  -H "Authorization: Bearer <YOUR_HF_TOKEN>"

gemini mcp add -t http huggingface https://huggingface.co/mcp?login

gemini extensions install https://github.com/huggingface/hf-mcp-server

Install in VSCode

Click here to add the Hugging Face connector directly to VSCode. Alternatively, install from the gallery at https://code.visualstudio.com/mcp:

If you prefer to configure manually or use an auth token, add the snippet below to your mcp.json configuration:

"huggingface": {
    "url": "https://huggingface.co/mcp",
    "headers": {
        "Authorization": "Bearer <YOUR_HF_TOKEN>"
    }

"huggingface": {
    "url": "https://huggingface.co/mcp",
    "headers": {
        "Authorization": "Bearer <YOUR_HF_TOKEN>"
    }

Once installed, navigate to https://huggingface.co/settings/mcp to configure your Tools and Spaces.

[!TIP] Add ?no_image_content=true to the URL to remove ImageContent blocks from Gradio Servers.

Quick Guide (Repository Packages)

This repo contains:

• (/mcp) MCP Implementations of Hub API and Search endpoints for integration with MCP Servers.
• (/app) An MCP Server and Web Application for deploying endpoints.

MCP Server

The following transports are supported:

• STDIO
• StreamableHTTP
• StreamableHTTP in Stateless JSON Mode (StreamableHTTPJson)

The Web Application and HTTP Transports start by default on Port 3000.

The StreamableHTTP service is available at /mcp. Although though not strictly enforced by the specification this is common convention.

[!TIP] The Web Application allows you to switch tools on and off. For STDIO and StreamableHTTP this will send a ToolListChangedNotification to the MCP Client. In StreamableHTTPJSON mode the tool will not be listed when the client next requests the tool lists.

Running Locally

You can run the MCP Server locally with either npx or docker.

npx @llmindset/hf-mcp-server       # Start in STDIO mode
npx @llmindset/hf-mcp-server-http  # Start in Streamable HTTP mode
npx @llmindset/hf-mcp-server-json  # Start in Streamable HTTP (JSON RPC) mode

To run with docker:

docker pull ghcr.io/evalstate/hf-mcp-server:latest
docker run --rm -p 3000:3000 ghcr.io/evalstate/hf-mcp-server:latest

All commands above start the Management Web interface on http://localhost:3000/. The Streamable HTTP server is accessible on http://localhost:3000/mcp. See [Environment Variables](#Environment Variables) for configuration options. Docker defaults to Streamable HTTP (JSON RPC) mode.

Developing OpenAI Apps SDK Components

To build and test the Apps SDK component, run

cd packages/app
npm run dev:widget

Then open http://localhost:5173/gradio-widget-dev.html. This will bring up a browser with HMR where you can send Structured Content to the components for testing.

Development

This project uses pnpm for build and development. Corepack is used to ensure everyone uses the same pnpm version (10.12.3).

# Install dependencies
pnpm install

# Build all packages
pnpm build

Build Commands

pnpm run clean -> clean build artifacts

pnpm run build -> build packages

pnpm run start -> start the mcp server application

pnpm run buildrun -> clean, build and start

pnpm run dev -> concurrently watch mcp and start dev server with HMR

Docker Build

Build the image:

docker build -t hf-mcp-server .

Run with default settings (Streaming HTTP JSON Mode), Dashboard on Port 3000:

docker run --rm -p 3000:3000 -e DEFAULT_HF_TOKEN=hf_xxx hf-mcp-server

Run STDIO MCP Server:

docker run -i --rm -e TRANSPORT=stdio -p 3000:3000 -e DEFAULT_HF_TOKEN=hf_xxx hf-mcp-server

TRANSPORT can be stdio, streamableHttp or streamableHttpJson (default).

Transport Endpoints

The different transport types use the following endpoints:

• Streamable HTTP: /mcp (regular or JSON mode)
• STDIO: Uses stdin/stdout directly, no HTTP endpoint

Stateful Connection Management

The streamableHttp transport is stateful - it maintains a connection with the MCP Client through an SSE connection. When using this transport, the following configuration options take effect:

Environment Variables

The server respects the following environment variables:

• TRANSPORT: The transport type to use (stdio, streamableHttp, or streamableHttpJson)
• DEFAULT_HF_TOKEN: ⚠️ Requests are serviced with the HF_TOKEN received in the Authorization: Bearer header. The DEFAULT_HF_TOKEN is used if no header was sent. Only set this in Development / Test environments or for local STDIO Deployments. ⚠️
• If running with stdio transport, HF_TOKEN is used if DEFAULT_HF_TOKEN is not set.
• HF_API_TIMEOUT: Timeout for Hugging Face API requests in milliseconds (default: 12500ms / 12.5 seconds)
• USER_CONFIG_API: URL to use for User settings (defaults to Local front-end)
• ALLOW_INTERNAL_ADDRESS_HOSTS: Optional comma-separated host allowlist to permit internal/reserved DNS resolutions for trusted domains during outbound checks (supports exact hosts and *. wildcards, for example: huggingface.co,*.hf.space).
• MCP_STRICT_COMPLIANCE: set to True for GET 405 rejects in JSON Mode (default serves a welcome page).
• AUTHENTICATE_TOOL: whether to include an Authenticate tool to issue an OAuth challenge when called
• SEARCH_ENABLES_FETCH: When set to true, automatically enables the hf_doc_fetch tool whenever hf_doc_search is enabled
• PROXY_TOOLS_CSV: Optional CSV that defines Streamable HTTP proxy tool sources (see below).
• GRADIO_SKIP_INITIALIZE: When set to true, Gradio MCP calls skip the initialize handshake and issue tools/call directly.

Proxy tools (Streamable HTTP via CSV)

You can load proxy tool definitions at startup by setting PROXY_TOOLS_CSV to a HTTPS URL or a local file path. The server fetches each MCP endpoint once on startup, runs initialize + tools/list (10s timeout), and registers any tools returned. If a source fails or returns no tools, it is skipped (no startup failure).

CSV format

proxy_id,url,response_type
papers,https://evalstate-hf-papers.hf.space/mcp,SSE
news,https://example.com/mcp,JSON

• proxy_id: identifier used to disambiguate tools.
• url: Streamable HTTP MCP endpoint.
• response_type: SSE (streamed response) or JSON (direct JSON-RPC response).

Tool naming

• Single source: tool names are unchanged (taken from the downstream server).
• Multiple sources: tool names are prefixed with proxy_id_ (e.g. papers_hf-papers-search_send).

You can include these tool names in bouquets or mixes as needed. Use bouquet=proxy or mix=proxy to enable all proxy tools loaded from PROXY_TOOLS_CSV (in

README truncated. View full README on GitHub.