GitHub MCP Server
GitHub's official MCP server, connecting AI agents directly to GitHub's platform to read repos, manage issues/PRs, analyze code and automate CI/CD workflows.
Overview
GitHub MCP Server is GitHub's own MCP implementation, giving AI agents and assistants direct, structured access to GitHub (repositories, issues, pull requests, Actions runs, code scanning alerts, discussions and more) through natural language instead of hand-rolled API calls.
How it works
The server ships in two forms: a remote server hosted by GitHub (the easiest way
to get started, with OAuth or a PAT) and a local server you run yourself via
Docker or a Go binary. Both expose the same GitHub capabilities grouped into
toolsets (repos, issues, pull_requests, actions, code_security,
discussions, projects, stargazers...), which you can enable selectively so the
agent's tool list stays small and focused.
- Connect: point any MCP-capable client (VS Code, Claude Code/Desktop, Cursor, Windsurf, Copilot CLI...) at the remote URL or the local Docker image.
- Pick toolsets: enable only what you need (
--toolsets repos,issues) orall. - Let the agent act: browse code, triage issues, open/review PRs, inspect Actions runs and security findings, all from chat.
- Read-only / lockdown modes: run with
--read-onlyto prevent writes, or--lockdown-modeto filter content from users without push access (reduces prompt-injection risk from untrusted issues/PRs).
Examples
Typical use cases straight from the project's own docs:
"Summarize open issues labeled 'bug' in this repo and suggest a triage order."
"Open a PR that fixes the failing lint step in the last Actions run."
"What Dependabot alerts are open on the main branch right now?"
Great for repo management, issue/PR automation, CI/CD intelligence and code/security analysis, anywhere an agent needs real GitHub context to act on rather than a guess.
Installation
For the remote server you need nothing local beyond an MCP client that supports HTTP servers. For the local server you need Docker.
- Easiest option: add the remote server URL
(
https://api.githubcopilot.com/mcp/) to your MCP client's config and sign in with OAuth or a personal access token when prompted. - Or run it locally with Docker instead:
and point your client at that process.docker run -i --rm -p 127.0.0.1:8085:8085 -e GITHUB_OAUTH_CALLBACK_PORT \ ghcr.io/github/github-mcp-server - Restart your client, then narrow the toolsets you actually need (for
example
--toolsets repos,issues) instead of enabling all of them. - For sensitive repos, add
--read-onlyor--lockdown-modeto limit what the agent can do.
Related assets
Awesome MCP Servers
The largest community catalog of MCP servers: thousands of entries across dozens of categories (databases, browser automation, cloud, communication and more), tagged by language and scope.
Context7
Pulls updated, version-specific documentation and code examples straight from the source into your prompt, so agents stop citing outdated APIs or hallucinating ones that don't exist.
Address Comments Agent
Works through pull request review comments one at a time: applies the minimal fix for each, pushes back on comments that don't make sense, adds test coverage, and commits with a descriptive message before moving to the next comment.