Install guide and config at curatedmcp.com

GitHub MCP: Manage Your Entire Repo Workflow from Your AI Agent

GitHub MCP is the official Model Context Protocol server that connects your AI agent directly to GitHub. Instead of switching between your editor, GitHub's web UI, and CI/CD dashboards, you can now handle pull requests, issues, workflows, and repository management through natural language commands in Claude, Cursor, or Windsurf.

What It Does

This server gives AI agents full visibility into your repositories and the ability to take action on them. Your agent can review pull requests with full code context, search and triage issues across multiple repos, query commit history and file changes, manage branch protections and collaborators, and trigger—or inspect—GitHub Actions workflows without leaving your development environment.

The practical upside: when you ask Claude to "review that PR and check if the tests passed," it actually opens the PR, reads the code, checks the CI logs, and gives you a detailed breakdown. You're not describing problems to your agent; your agent is seeing them directly.

Features include creating and merging PRs, searching issues and comments, managing collaborators, accessing security alerts and dependency insights, and fetching repository metadata for code browsing. It's the kind of integration that turns your AI agent from a chat interface into an actual development teammate.

How to Install

Install the MCP server via npm:

npx -y @github/mcp

Then add it to your Claude Desktop configuration:

{
  "mcpServers": {
    "github-mcp": {
      "command": "npx -y @github/mcp"
    }
  }
}

Authenticate with your GitHub token when prompted, and you're ready to go.

Real-World Use Cases

• Automated PR Triage: Ask Claude to "find all open PRs with failing tests, summarize the failures, and suggest fixes." Your agent reads each PR, checks the workflow logs, and delivers actionable feedback in seconds.
• Issue Investigation: "Search for all open bugs labeled 'critical' and create a summary of what needs fixing." The agent finds issues across repos, reads comments and linked PRs, and synthesizes context.
• Workflow Debugging: When a GitHub Actions pipeline fails, ask your agent to "check the latest workflow run for my deploy job and show me what went wrong." It fetches the logs directly and explains the error.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.