Skip to content
AI Dev Toolkit.
Esc
  • AI Token CounterCount tokens for GPT, Claude, Gemini, DeepSeek, Qwen and more.Tool
  • LLM API Cost CalculatorEstimate per-request, daily and monthly API costs.Tool
  • AI Model ComparisonCompare prices, context windows and features across models.Tool
  • AI Model Pricing PagesSpecs, real costs and cheaper alternatives for popular models.Tool
  • Context Window CheckerSee whether your text fits each model’s context window.Tool
  • Subscription vs API CalculatorFind out whether a chat plan or the API is cheaper for you.Tool
  • GPU / VRAM CalculatorCheck how much VRAM a local model needs and which GPUs fit.Tool
  • Prompt Caching CalculatorEstimate savings from prompt caching.Tool

MCP server · AI and machine learning

Hugging Face MCP server

Lets an assistant search Hugging Face models, datasets, Spaces and papers, run Jobs and call Gradio apps as tools.

Maintainer
Hugging Face (official)
Licence
MIT
Runs
Remote (URL)
Sign-in
OAuth sign-in in the app
Registry version
0.2.33 (updated 2025-10-22)
Last activity
2026-10-13

Overview

What it does

Hugging Face hosts its official MCP server at huggingface.co/mcp. Adding ?login to the URL starts an OAuth sign-in; you can instead send a Hugging Face token in an Authorization header. Once connected, the assistant can search and explore the Hub and call MCP-compatible Gradio apps hosted on Spaces. You choose extra tools and Spaces on your MCP settings page at huggingface.co/settings/mcp. The source is open, and the README also documents running it locally with npx or Docker.

Use it when you research models or datasets, look up library documentation, or want to call a hosted AI app, such as image generation or transcription, from your assistant. A built-in hf_fs tool handles most Hub navigation, including semantic search of docs and Spaces. Optional tools create repositories and write files to them, run and schedule Jobs on Hugging Face infrastructure, and manage sandboxes. A Dynamic Spaces option lets the assistant find and use MCP Spaces on the fly.

Tools

What the assistant can do

  • Search models, datasets, Spaces and papers on the Hub
  • Semantic search of Hugging Face documentation
  • Navigate the Hub with the hf_fs tool
  • Call MCP-compatible Gradio Spaces as tools
  • Run, monitor and schedule Jobs
  • Create and use sandboxes
  • Create repositories and write files to them

Install

Install Hugging Face MCP server in your app

You need:

  • A Hugging Face account for sign-in and for repository, Job, dynamic Space and sandbox tools
  • An MCP client with remote server support

Run in your project. --scope local keeps the servers private to you in this project; use user for every project.

Terminal
claude mcp add --scope local --transport http hugging-face 'https://huggingface.co/mcp?login'
  • hugging-face: if the server needs a sign-in, the app will ask (run /mcp in Claude Code).

Want it alongside other servers? Open Hugging Face MCP Server in the MCP config generator and add the rest; it writes one file per app. The snippets follow the setup in the server’s README, checked on 2026-10-14.

Before you install

  • Repository, Job, dynamic Space and sandbox tools act with your Hugging Face account, so they can create repos, write files and start jobs.
  • Community Spaces are third-party apps. Arguments the assistant sends to a Space tool go to that Space, so add only Spaces you trust.
  • If you use a token header instead of OAuth, give the token only the access the tools you enable need.
  • A remote server sees whatever the assistant sends it. Sign in with the narrowest access the service offers. See how to choose a safe server.

Why it’s listed

Why we list it

The official gateway to the largest open model hub, with hosted setup and access to thousands of Gradio apps as tools.

Related

All MCP serversMCP config generatorGuide: MCP servers in Claude CodeGuide: what is MCP?