Hugging Face MCP Course certificate

Learning plan

About 15 hours of focused study, split into 3 phases. Pick your weekly pace and the timeline adjusts.

Estimated finish: 2 weeks

What the exam covers

Source: official exam guide. Weights can change, so check before you start.

Before you start

• Experience with Python or TypeScript and calling HTTP APIs. The Hugging Face Agents Course helps but is not required
• A free Hugging Face account, Node.js (for the MCP Inspector) and an MCP-capable client such as Claude Desktop, Claude Code or Cursor

1. Week 1 · 4 h

   MCP architecture and the fundamentals certificate (Unit 1)

   • Describe the host, client and server roles and the JSON-RPC message flow (initialize, list, call, notifications)
   • Distinguish the three server primitives (Tools, Resources, Prompts) and when to use each
   • Compare stdio and Streamable HTTP transports and when each fits
   • Pass the Unit 1 quiz to earn the fundamentals certificate

   Use: MCP Course – Unit 1 · Unit 1 certificate quiz · Model Context Protocol documentation

2. Week 1 · 6 h

   Build and deploy an end-to-end MCP app (Unit 2)

   • Expose a Python function as an MCP tool using Gradio's mcp_server option, and deploy it to a Hugging Face Space
   • Connect to it from a smolagents Python client and from a Hugging Face JS client
   • Inspect tool schemas and calls with the MCP Inspector

   Hands-on: Build the Unit 2 sentiment-analysis MCP server in Gradio, deploy it to a free Space, then add a second tool of your own (for example keyword extraction). Call both from Claude Desktop or Cursor and check the calls in the MCP Inspector.

   Use: MCP Course – Unit 2 · Gradio guide to building an MCP server · MCP Inspector (GitHub)

3. Week 2 · 5 h

   A workflow server for Claude Code (Unit 3)

   • Build the PR Agent Workflow Server, which suggests PR templates from changed files
   • Receive GitHub Actions results through webhooks over a Cloudflare Tunnel and expose them as MCP Prompts
   • Send Slack notifications on deployment results using Tools, Resources and Prompts together

   Hands-on: Complete the Unit 3 server against a real GitHub repo of your own. Trigger a failing CI run and confirm Claude Code can summarise the failure through your MCP prompt.

   Use: MCP Course – Unit 3

You're ready to book when…

• You hold the Unit 1 fundamentals certificate
• You can explain when to model something as a Tool, a Resource or a Prompt, with an example of each
• You have a deployed MCP server that a third-party client can call successfully

Exam tips

• The fundamentals certificate comes from the Unit 1 quiz. If the embedded quiz doesn't load, open the quiz Space on the Hub directly.
• The completion certificate needs the use-case units (2 and 3), so keep your Space and repo public so you can show the work.