Hugging Face

Post image

Try a prompt

A visualization of what the Hugging Face MCP server lets an agent do. This page reads the public Hub API. It does not call the MCP server.

Connect

One endpoint

The hosted server is huggingface.co/mcp. Open huggingface.co/settings/mcp while logged in, pick a client (Cursor, VS Code, Claude, and others), and paste the snippet that page generates. The docs say to use that page rather than writing config by hand. The public setup page’s OAuth URL is https://huggingface.co/mcp?login. Manual Cursor config with a read token:

mcp.json

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

Replace <HF_TOKEN> with a Hugging Face read token. That manual block is the one shown on huggingface.co/mcp for Cursor.

Catalog

Tools at a glance

Names as the docs list them. Extra tools are turned on from the MCP settings page. Experimental tools are added from time to time.

`hf_fs`

Navigate the Hub, including semantic searches of Documents and Spaces. The docs say most Hub tasks can be done with this built-in tool.

Contribute Repos

Create and write files to those created repositories.

Sandboxes

Create and use Sandboxes: includes Sandbox file management.

Run and Manage Jobs

Run, monitor, and schedule jobs on Hugging Face infrastructure.

Sources

Hugging Face’s pages

Hugging Face MCP Server — what it does, prompts, and built-in tools
huggingface.co/mcp — hosted endpoint and client setup
MCP settings — client snippet, Spaces, and extra tools

From this series

Richard's results

Coming soon.

Richard will add what he tries himself. Nothing here yet.

← Back

#MCPMonday

Post image

Live view

The square is the image for the post. The wide frame is the same constellation at 1200×627. Both are drawn from Hub repos, not from an MCP call.