GitHub Dev Copilot - MCP Server

A production-ready Model Context Protocol (MCP) server that acts as a GitHub Copilot for AI agents. It provides tools to interact with the GitHub API, allowing agents to manage repositories, issues, PRs, and analyze repository health.

Architecture

+----------------+      +-----------------------+      +-------------+
|    AI Agent    | ---> |  GitHub MCP Server    | <--> |  GitHub API |
| (Client)       | STDI |  (Python / FastMCP)   | HTTP |             |
+----------------+      +-----------------------+      +-------------+
                                   |
                        +-----------------------+
                        |        Tools          |
                        | - list_repositories   |
                        | - read_issues         |
                        | - create_issue        |
                        | - comment_on_pr       |
                        | - analyze_repo_health |
                        +-----------------------+

Related MCP server: Narad GitHub Agent

Features

• Authentication: Securely connects using a GitHub Personal Access Token (PAT).

• Tooling: Exposes robust, strictly-typed MCP tools.

• Safety:

  • Validates inputs using Pydantic.

  • Rate limits API calls to prevent blocking.

  • Never exposes tokens.

• Analysis: intelligent analyze_repo_health tool to score repositories.

Setup

1. Clone the repository (if applicable) or navigate to the project directory.

2. Install Dependencies:

   pip install -r requirements.txt

3. Configure Environment: Copy .env.example to .env and add your GitHub PAT.

   cp .env.example .env
   # Edit .env and set GITHUB_TOKEN

4. Run the Server: You can run it directly with Python, or via an MCP client inspector.

   python main.py

Using with Claude Desktop or MCP Clients

Add the server config to your claude_desktop_config.json:

{
  "mcpServers": {
    "github-copilot": {
      "command": "python",
      "args": ["/path/to/mcp_github_dev_copilot/main.py"]
    }
  }
}

Available Tools

1. list_repositories

Lists repositories accessible to the authenticated user.

• Returns: Name, owner, private/public status, stars, language.

2. read_issues

Fetches issues from a repository.

• Input: repo_owner, repo_name, state ("open", "closed", "all")

• Output: List of issues with ID, title, labels.

3. create_issue

Creates a new issue.

• Input: repo_owner, repo_name, title, body (markdown), labels

• Output: Issue URL and number.

4. comment_on_pr

Adds a comment to a Pull Request.

• Input: repo_owner, repo_name, pr_number, comment_text

• Output: Comment URL.

5. analyze_repo_health

Performs a health check on a repository.

• Checks: Open issues count, stale issues (>30 days), recent commit activity, tests presence, CI config.

• Output: Health score (0-100) and improvement suggestions.

Security & Limitations

• Rate Limiting: The server implements a decorator to handle GitHub rate limits by waiting or retrying.

• Scopes: Ensure your PAT has repo, read:org, and workflow scopes.

• No Browser: This server uses the official GitHub API, not web scraping.

• Logs: Activities are logged to stderr to avoid interfering with MCP stdio communication.

Future Improvements

• Add update_issue and merge_pr tools.

• Implement file content reading/searching.

• Add support for GitHub App authentication.