🦙 GitHub MCP Agent with LLaMA 3

Explore any public GitHub repository using natural language with the power of Meta's LLaMA 3 and Model Context Protocol (MCP).

Built with:

• 🦙 LLaMA 3 8B Instruct via Hugging Face
• 🧠 Model Context Protocol (MCP) for GitHub insights
• 🎈 Streamlit for an interactive UI
• 🔧 Agno Framework for agent management

!GitHub MCP Agent Screenshot

---

🚀 Features

• 🔍 Ask natural language questions about any GitHub repository
• 📊 Get live insights using MCP: issues, PRs, metrics, trends
• 🤖 Answers powered by LLaMA 3's reasoning and summarization
• 🧰 Fully local + API-based, uses your GitHub token securely

---

🧩 App Flow

Here’s a breakdown of how the app works:

1. Authentication:

   • When you first open the app, you’ll need to provide your GitHub token in the sidebar.
   • The token allows the app to access private/public repositories on your behalf. (Create one at GitHub Token Settings)

2. Repository and Query Input:

   • You can enter the GitHub repository (e.g., facebook/react) that you want to analyze.
   • You can also choose a predefined query type (such as Issues, Pull Requests, or Repository Activity) or type your custom query.

3. Query Processing:

   • After you click "Run Query", the app sends your prompt to a powerful LLaMA 3 8B Instruct model hosted via Hugging Face.
   • The query is processed using both the LLaMA 3 model and Model Context Protocol (MCP) to fetch data about the repository from GitHub (like issues, PRs, repo activity).

4. Results Display:

   • The results are then presented in a streamlit UI, formatted neatly with markdown, tables, and links to GitHub.
   • You can see insights like open issues, PR statuses, recent repository activity, and more.

5. Interactive Exploration:

   • You can continue interacting by modifying the query, selecting a different repo, or using different query types.

---

📦 Installation

https://github.com/sandy1279/gitbot-mcp
cd gitbot-mcp
pip install -r req.txt