GitHub Repo Q&A Agent – Docker MCP Toolkit Example

This repository demonstrates a real-world developer setup for building an AI agent with the Docker MCP Toolkit that can connect to a GitHub repository and answer questions about its contents.

The agent:

• Summarizes repositories
• Explains functions and modules
• Finds where specific API calls are made
• Retrieves code references with file/line details

It uses:

• Docker MCP Toolkit for tool integration (GitHub API)
• Docker Compose for orchestration
• OpenAI GPT-4o for reasoning and natural language responses

---

Prerequisites

• Docker & Docker Compose
• Python 3.10+
• OpenAI API key
• GitHub Personal Access Token with at least:
  • repo
  • read:org (for private repos)

---

Quick Start

1. Clone the repository

git clone https://github.com/<your-username>/<your-repo>.git
cd <your-repo>

2. Start the MCP Gateway

docker compose up -d mcp-gateway-github

3. Configure GitHub Access

docker mcp secret set github.personal_access_token=<YOUR_GITHUB_TOKEN>

4. Create .env for OpenAI

Create a file named .env in the root directory and add your OpenAI API key:

OPENAI_API_KEY=sk-********************************

5. Install dependencies

Create and activate a Python virtual environment, then install dependencies from requirements.txt:

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.tx

6. Run the agent

python app.py

7. Ask questions

Example usage:

Enter your query: Summarize https://github.com/owner/repo

Extending the Setup

• Add more MCP connectors (e.g., Jira, Slack, Databases) by adding new services to docker-compose.yml.
• Containerize the agent itself for a fully portable deployment.
• Integrate with CI/CD pipelines using the same Compose file.