Connect GitHub MCP to repositories, issues and pull requests with GitHub’s hosted service or a locally run server.
GitHub’s official MCP server connects an assistant to repository content and development workflows. Its tools cover issues, pull requests and other GitHub resources, with the available operations determined by the selected toolsets and the authenticated account.
Start with a repository and issue URL. Ask the assistant to inspect the issue, find the relevant files and explain the connection using file paths or commit references. A follow-up can prepare a pull request or comment, but the destination and proposed change should be explicit. An account with access to one repository does not automatically have access to another.
The hosted endpoint is https://api.githubcopilot.com/mcp/. OAuth availability depends on the client’s GitHub integration; a personal access token is another documented route. The local server runs through Docker or a release binary and now also documents OAuth, alongside token authentication. Follow the current client guide rather than assuming that every local installation requires a personal access token.
claude mcp add --transport http github https://api.githubcopilot.com/mcp/Use the README’s client-specific authentication instructions; GitHub OAuth support is not identical across clients.
A GitHub account with access to the repositories you need. Use a client with the documented OAuth integration, or follow the token authentication instructions.
| Variable | Description | Required |
|---|---|---|
| GITHUB_PERSONAL_ACCESS_TOKEN | PAT for local mode (remote mode uses OAuth instead)e.g. github_pat_... | Optional |
| GITHUB_TOOLSETS | Comma-separated toolsets to enable (e.g. repos,issues,pull_requests)e.g. repos,issues,pull_requests,actions | Optional |
Read, search and edit files inside allowed directories with the Model Context Protocol reference Filesystem server.
Retrieve version-aware library documentation and code examples with Context7 MCP, using resolve-library-id and query-docs.
Free catalog for Fabler's USDC-paid x402 agent tools on Base; no account or wallet required.
PlanExe exposes plan creation, progress inspection and recovery workflows through a hosted or self-hosted MCP service.
This Model Context Protocol (MCP) server enables AI assistants and applications to:
An LLM-based code review MCP server with AST-powered smart context extraction