🚀 GitHub AI DevOps Agent (MCP + Gemini)

An AI-powered DevOps assistant that interacts with GitHub using natural language.

Built using:

• LangChain + LangGraph
• Google Gemini API
• Model Context Protocol (MCP)
• Custom GitHub tools

---

✨ Features

• 🧠 Supports natural language prompts (English)
• 📦 Create public/private repositories
• 🌿 Create branches
• 📄 Create/update files (README, etc.)
• 🔍 List repositories, branches, commits
• 🐛 Create issues & PRs
• 💬 Comment on issues
• 🗑️ Delete repositories (custom tool with safety confirmation)
• 🔁 Retry & fallback logic
• 🔐 Safe execution for destructive actions

---

✨ What This Project Does

Control GitHub like this:

"Create a branch named feat/aiagent-test in your repository (replace <reponame> with your repo name)"

Example:

Create a branch named feat/aiagent-test in ai-devops-demo

The agent will:

• Understand your intent
• Select the correct GitHub tool
• Execute it automatically

No GitHub API knowledge required.

---

🏗️ Architecture

User Input
   ↓
AI Agent (LangGraph ReAct)
   ↓
Gemini (LLM - reasoning & decision making)
   ↓
Tools Layer
   ├── MCP GitHub Server (@modelcontextprotocol/server-github)
   └── Custom Tools (delete_repository)
   ↓
GitHub API

---

🧑‍💻 One-Time Setup (New Laptop Friendly)

Follow these steps exactly — after this, the project will run without issues.

---

1️⃣ Install System Dependencies

Install Python (>= 3.10)

python --version

If not installed → https://python.org

---

Install Node.js (Required for MCP)

Download: https://nodejs.org

node -v
npm -v

---

Install Git

git --version

---

2️⃣ Get Required Credentials

Gemini API Key

• https://aistudio.google.com → Create API key

GitHub Personal Access Token

Scopes required:

• repo
• workflow
• read:org

GitHub Username

Example:

iam-pavanpachuru

---

3️⃣ Clone the Repository

git clone https://github.com/iam-pavanpachuru/github-ai-agent
cd github-ai-agent

---

4️⃣ Create Virtual Environment

python -m venv venv

Activate:

Linux / Mac

source venv/bin/activate

Windows

venv\\Scripts\\activate

---

5️⃣ Install Python Dependencies

pip install --upgrade pip
pip install -r requirements.txt

---

6️⃣ Setup MCP (Node Environment)

Fix npm permissions (Linux/Mac)

mkdir -p ~/.npm-global
npm config set prefix ~/.npm-global
echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc
source ~/.bashrc

---

Install MCP Server

npm install -g @modelcontextprotocol/server-github

---

Verify MCP Works

npx @modelcontextprotocol/server-github

Expected:

GitHub MCP Server running on stdio

---

7️⃣ Configure Environment Variables

Create .env file:

GOOGLE_API_KEY=your_gemini_api_key
GITHUB_PERSONAL_ACCESS_TOKEN=your_github_token
GITHUB_USERNAME=your_github_username  # e.g., iam-pavanpachuru

Protect it:

echo ".env" >> .gitignore

---

🔌 MCP Usage Modes (Important)

✅ Recommended — Auto Mode

• Python reads .env
• Passes token to MCP internally
• No manual setup required

---

⚙️ Manual Mode

export GITHUB_PERSONAL_ACCESS_TOKEN=your_token
npx @modelcontextprotocol/server-github

---

🚀 Run the Project

python github_ai_agent.py

Type your task and press ENTER twice.

---

🧪 Validate Setup (Important)

Before running real tasks, verify everything is working correctly.

1. Validate Environment Variables

python test_mcp_client.py

Expected:

• MCP connects successfully
• GitHub authentication works

---

2. Validate MCP Tools

python inspect_mcp_tools.py

Expected:

• List of available GitHub tools
• Confirms MCP server is working

---

3. Quick Test Prompt

Run the agent and try:

List all repositories for my GitHub account

If this works → your setup is fully correct ✅

---

🧪 Example Prompts (MCP Tools)

Here are some example prompts demonstrating how the agent interacts with GitHub using MCP tools:

---

📦 Create Repository

Create a new private repository called ai-devops-demo
If empty, create a README.md with content "# AI DevOps"

---

📄 Create or Update File

Create a README.md file in ai-devops-demo with content "# AI DevOps Project"

---

🌿 Create Branch

Create a new branch called feat/auth in repository ai-devops-demo

---

🔍 Search Repositories

Search for repositories related to "devops automation"

---

🐛 Create Issue

Create an issue in ai-devops-demo with title "Bug: Fix login error" and description "Login fails on invalid credentials"

---

🔀 Create Pull Request

Create a pull request from branch feat/auth to main in ai-devops-demo with title "Add authentication feature"

---

💬 Add Comment to Issue

Add a comment to issue #1 in ai-devops-demo saying "Working on this issue"

---

🔎 List Commits

List all commits from the main branch in ai-devops-demo

---

🗑️ Delete Repository

Delete the repository ai-devops-demo

⚠️ Requires confirmation before execution.

---

🔧 Custom Tool: Delete Repository

The MCP GitHub server does not support repository deletion.

So we added a custom tool:

def delete_github_repo(repo_name: str):

This directly calls GitHub REST API:

DELETE /repos/{owner}/{repo}

---

📁 Project Structure

.
├── github_ai_agent.py      # Main AI agent
├── inspect_mcp_tools.py   # Utility to inspect MCP tools
├── test_mcp_client.py     # MCP connection testing
├── .env                   # Environment variables
├── .gitignore
└── venv/

---

📌 File Explanations

🔹 github_ai_agent.py

Main application:

• Handles user input
• Connects to MCP server
• Loads tools
• Runs AI agent
• Executes GitHub operations

---

🔹 inspect_mcp_tools.py

Purpose: Tool discovery & debugging

Run:

python inspect_mcp_tools.py

Helps you:

• See all available MCP tools
• Understand system capabilities
• Debug why certain prompts fail

---

🔹 test_mcp_client.py

Purpose: Testing & validation

Used for:

• Verifying MCP connection
• Testing GitHub authentication
• Running quick experiments

---

⚠️ Known Limitations

• Gemini free tier has strict rate limits (429 errors)
• Some MCP tools may be missing (e.g., delete repo)
• Model availability may change (503 errors)

---

🧠 Key Learnings

• LLMs don’t execute actions — tools do
• MCP acts as a bridge between AI and real systems
• Safety checks are critical for destructive operations
• Prompt clarity improves tool usage

---

🔮 Future Improvements

• Will be updated with more features and enhancements soon 🚀

---

🤝 Contributing

Feel free to fork, improve, and share!

---

⭐ If you like this project

Give it a star ⭐ and share on LinkedIn 🚀

---

👨‍💻 Author

Pavan Kalyan Pachuru

• GitHub: https://github.com/iam-pavanpachuru
• LinkedIn: https://www.linkedin.com/in/pavan-kalyan-pachuru-538a4016b