ISLAM-PO/mcp-pr-agent

MCP PR Agent — Model Context Protocol Course Project

MCP Course (Units 1–3) project for the Hugging Face MCP Course, built in partnership with Anthropic.

A production-ready MCP server that helps developers create better pull requests by analyzing git changes and suggesting appropriate PR templates. The server exposes 3 tools, 2 resources, and 1 prompt via the Model Context Protocol.

Table of Contents

• Overview
• Project Structure
• Tools
• Quick Start
• Usage with Claude Code
• Quiz Answers (Unit 1)
• Certifications
• License

Overview

This project demonstrates the full MCP workflow:

1. Unit 1 (Fundamentals): MCP architecture (Host/Client/Server), three capability types (Tools/Resources/Prompts), transport mechanisms (stdio/HTTP+SSE), JSON-RPC 2.0
2. Unit 2 (Building with MCP): Implementing MCP clients and servers using SDKs
3. Unit 3 (Deployed Use Case): Building a deployed MCP application (PR agent with GitHub/Slack integration)

Project Structure

mcp_course_project/
├── README.md
├── LICENSE
├── requirements.txt
├── mcp-server/
│   └── server.py          # MCP PR agent server (3 tools, 2 resources, 1 prompt)
├── templates/
│   ├── bug.md
│   ├── feature.md
│   ├── docs.md
│   ├── refactor.md
│   ├── test.md
│   ├── performance.md
│   └── security.md
└── quiz-answers/
    └── QUIZ_ANSWERS.md    # Unit 1 quiz (5/5 correct)

Tools

Tool	Description
analyze_file_changes	Get git diff, changed files, commits, and statistics. Supports truncation for large diffs.
get_pr_templates	List all available PR templates with their content.
suggest_template	Suggest the most appropriate PR template based on change type.

Resources

Resource	Description
templates://list	List all available PR templates
config://project	Project configuration and metadata

Prompts

Prompt	Description
review_code	Generate a code review prompt for a given code snippet

Quick Start

1. Install dependencies

pip install -r requirements.txt

2. Run the MCP server

python mcp-server/server.py

The server uses stdio transport by default, compatible with Claude Code, Cursor, and other MCP clients.

3. Test with Claude Code

# Add the MCP server
claude mcp add pr-agent -- python /absolute/path/to/mcp-server/server.py

# Verify it's configured
claude mcp list

# Use it
# "Analyze my changes and suggest a PR template"

Usage with Claude Code

Once configured, you can ask Claude:

• "Analyze my file changes and suggest a PR template"
• "What PR templates are available?"
• "Review this code for issues"

Quiz Answers (Unit 1)

All 5 quiz questions with correct answers are in quiz-answers/QUIZ_ANSWERS.md.

Key concepts:

• Host: User-facing AI application (Claude Desktop, Cursor)
• Client: Manages 1:1 communication with an MCP Server
• Server: Exposes tools, resources, and prompts via MCP
• M×N Problem: M apps × N tools = M×N integrations → MCP reduces to M + N

Take the interactive quiz: https://huggingface.co/spaces/mcp-course/unit_1_quiz

Certifications

Certificate	Requirements	Link
Fundamentals	Complete Unit 1 (pass quiz ≥80%)	https://huggingface.co/spaces/mcp-course/unit_1_quiz
Completion	Complete Units 2 & 3 (build + deploy MCP app)	Project-based (this repo demonstrates Unit 3)

License

MIT License. See LICENSE.

Acknowledgements

Built as the Unit 3 project of the Hugging Face MCP Course, built in partnership with Anthropic. Source starter code from huggingface/mcp-course.