GitHub MCP CLI with Ollama

A terminal-based project that uses the Model Context Protocol (MCP) and local LLMs (via Ollama) to query and summarize GitHub data.

Overview

This project is split into two components:

1. MCP Server (server.py): Built with fastmcp, it provides tools to fetch GitHub repository details, recent issues, and recent commits.

2. CLI Client (client.py): An interactive terminal interface using rich and LangChain that talks to a local Ollama model to interact with the MCP tools.

Related MCP server: GitHub MCP Server

Setup

1. Install dependencies:

   pip install -r requirements.txt

2. Configure GitHub Access: For increased rate limits and access to private repositories, export your GitHub Personal Access Token (PAT) as an environment variable (or put it in a .env file):

   export GITHUB_TOKEN="your_personal_access_token_here"

3. Start Ollama: Ensure Ollama is running locally with a tool-capable model (like llama3.2):

   ollama run llama3.2

4. Run the client:

   python client.py

Development

Currently, both server.py and client.py are fully operational. The client uses MultiServerMCPClient and LangChain's React Agent to seamlessly connect to the tools and route your questions to Ollama.

Tools exposed by the MCP Server:

• get_repo(owner, repo): Get basic details about a repository.

• list_issues(owner, repo, state, limit): Fetch recent issues.

• list_commits(owner, repo, limit): Fetch recent commits.