🤖 DevFlow AI — Claude-Powered Pull Request Automation

A fully-integrated MCP (Model Context Protocol) server that turns pull request chaos into team-wide clarity.
This agent analyzes code changes, monitors CI/CD pipelines, and sends real-time Slack updates — all powered by Claude Code and the MCP standard.

---

🚀 Features

• 🧠 AI-Powered PR Suggestions
  Uses Claude Code to analyze git diffs and suggest the best PR templates with clear, actionable descriptions.

• 🧪 CI/CD Monitoring (GitHub Actions)
  Tracks workflows via GitHub webhooks, highlights failures, and summarizes deployment outcomes.

• 📣 Slack Team Notifications
  Sends professionally formatted alerts and summaries using Slack markdown and team routing logic.

• 💠 Plug-and-Play Tools & Prompts
  Designed using MCP primitives:

  • Tools for gathering structured data (git, CI, Slack)
  • Prompts for guiding Claude through consistent workflows
  • Resources for project context and team conventions

---

🔍 Why This Matters

Development teams often struggle with:

• Poor pull request descriptions
• Missed CI/CD failures
• Communication silos across teams

This project addresses all three. It automates tedious dev workflows while preserving human context and clarity — so your team ships confidently, communicates clearly, and spends less time deciphering PRs or tracking down failures.

---

📆 Technologies Used

• Claude Code + Model Context Protocol (MCP)
• Python 3.11+
• FastMCP — for MCP server scaffolding
• GitHub Webhooks — via aiohttp and Cloudflare Tunnel
• Slack Webhooks — with secure environment-based secrets
• Git CLI — for accurate diff and commit history analysis

---

🧹 Architecture Overview

Developer → Git Push → GitHub → Webhook Server
                                  ↓
      Claude Code ← MCP Server ← github_events.json
            ↓                     
    Slack Notification

Component	Description
server.py	The main MCP server with all Tools & Prompts
webhook_server.py	Listens for GitHub Actions events
github_events.json	Stores the last 100 CI/CD events locally
Templates	Structured natural language report templates for Claude

---

⚙️ Getting Started

🔧 Prerequisites

• Python 3.11+
• Claude Code CLI
• GitHub repo with Actions enabled
• Slack workspace with incoming webhooks
• uv package manager (or use pip/venv as needed)

🧲 Setup & Run

# Clone the repo
git clone https://github.com/your-username/devflow-ai.git
cd devflow-ai

# Install dependencies
uv venv .venv
source .venv/bin/activate
uv sync

# Start webhook listener 
python webhook_server.py

# Start MCP server (in separate terminal)
python server.py

# Tunnel webhook to GitHub
cloudflared tunnel --url http://localhost:8080

🔐 Environment Configuration

# Set your Slack Webhook URL securely
export SLACK_WEBHOOK_URL="https://hooks.slack.com/services/XXXX/YYYY/ZZZZ"

---

✅ Example Usage

Ask Claude:

"Can you analyze my code changes and suggest a PR template?"
"Summarize our recent deployment and notify the team in Slack."
"Troubleshoot the failed ci-lint GitHub Action."

Claude will:

• Call your registered MCP Tools
• Use project Resources for context
• Apply your Prompts to return consistent reports

---

📈 Impact

Before	After
PRs with titles like "fix" and "update stuff"	Auto-generated summaries with correct templates and security checks
Developers manually checking Actions tab	Real-time notifications with status breakdowns
Missed cross-team updates	Slack alerts aligned with team preferences
Repeated debugging of known issues	Prompt-based guidance from recent history

---

💪 Testing

Run unit tests:

uv run pytest tests/unit/

Run integration tests:

uv run pytest tests/integration/

Use manual_test.md to simulate GitHub events locally via curl.

---

📋 Learn More

• MCP Documentation
• Claude Code Guide
• Slack Message Formatting
• GitHub Webhooks Guide