cap-mcp

MCP (Model Context Protocol) server for querying indexed GitHub repositories and git histories.

Overview

This MCP server provides access to an AI-powered knowledge base of indexed GitHub repositories and git histories. Ask questions about code, commits, pull requests, issues, and repository structures to get intelligent context-aware answers.

Installation

From npm (once published)

npm install -g cap-mcp

From source

Clone the repository and build:

git clone https://github.com/yourusername/cap-mcp.git
cd cap-mcp
npm install
npm run build

Usage

The MCP server provides a get_ai_context tool that queries indexed GitHub repositories and git histories.

Running the MCP Server

npm run dev

Configure in Cursor

Add this to your Cursor MCP config file at ~/.cursor/mcp.json:

Recommended (auto-downloads latest from npm):

{
  "mcpServers": {
    "cap-mcp": {
      "command": "npx -y cap-mcp@latest",
      "env": {}
    }
  }
}

• @latest (Optional) ensures you always get the newest version
• -y automatically installs without prompting for confirmation

Configure in Claude Desktop

Add this to your Claude Desktop MCP config file at ~/.claude/mcp.json (or %APPDATA%\Claude\mcp.json on Windows):

{
  "mcpServers": {
    "cap-mcp": {
      "command": "npx -y cap-mcp@latest",
      "env": {}
    }
  }
}

Note: The configuration is identical for both Cursor and Claude Desktop since they both use the same Model Context Protocol standard.

Available Tools

• get_ai_context: Get context from indexed GitHub repositories and git histories
  • query: Your question about the repositories or git history (required)
    • Examples: "How does authentication work?", "What changed in the last commit?", "Show me all API endpoints"
  • additional_context: Additional context like repository name, file paths, branch names, commit SHAs, or date ranges (optional)

Use Cases

• 🔍 Code Search: "Where is the user authentication implemented?"
• 📝 Commit History: "What changes were made to the API in the last week?"
• 🔧 Code Understanding: "How does the payment processing work?"
• 🌳 Repository Structure: "What are the main components of this project?"
• 🐛 Issue Tracking: "What issues are related to database performance?"
• 🔀 Pull Request Context: "Summarize recent pull requests for the auth module"

Backend Requirements

The MCP server connects to an AI endpoint at http://127.0.0.1:5000/api/agent that:

• Has indexed GitHub repositories and git histories
• Accepts POST requests
• Receives JSON body with: { "message": "...", "stream": false }
• Returns JSON response with context from the indexed repositories

Example Queries

Query: "How does the authentication system work?"
Additional Context: "repository: my-app, focus on backend"

Query: "What files were changed in commit abc123?"
Additional Context: "repository: my-app"

Query: "Show me all REST API endpoints"
Additional Context: "repository: my-api, path: src/routes"

Query: "What are the recent changes to the database schema?"
Additional Context: "repository: my-app, last 30 days"