Best for
- Teams that want an AI assistant to directly operate on GitHub repos and collaboration workflows
- Users already in the GitHub Copilot ecosystem who want a zero-deployment remote option
G
GitHub MCP Server
Official
GitHub's official MCP server for managing repos, issues, PRs, and workflows via natural language
GitHub source repository ↗
76FMRS · B
Maintained by the official GitHub team; toolsets cover the full repository lifecycle. Both local Docker and remote HTTP paths are clearly documented. This review is based on the README and server.json, not an actual handshake test — verify with a read-only toolset before enabling write access.
The official GitHub-maintained MCP server, exposing GitHub's core capabilities (repositories, issues, pull requests, Actions, code security scanning, Discussions, orgs and projects) through grouped toolsets. Runs locally via Docker, or connects directly to GitHub's hosted remote Streamable HTTP endpoint (api.githubcopilot.com/mcp/) with no local install required.
Tools
context
Tools that provide context about the current user and GitHub context (strongly recommended)
actions
GitHub Actions workflows and CI/CD operations
code_quality
GitHub Code Quality related tools
code_security
Code security related tools, such as GitHub Code Scanning
discussions
GitHub Discussions related tools
gists
GitHub Gist related tools
git
GitHub Git API related tools for low-level Git operations
issues
GitHub Issues related tools
labels
GitHub Labels related tools
notifications
GitHub Notifications related tools
orgs
GitHub Organization related tools
projects
GitHub Projects related tools
pull_requests
Create, review, and manage pull requests
repos
Repository management tools
users
User lookup tools
Setup
1) Local: docker pull ghcr.io/github/github-mcp-server, then start it with the config above — first run triggers an OAuth login, or supply a GITHUB_PERSONAL_ACCESS_TOKEN env var to skip it.
2) Remote: point your client's URL at https://api.githubcopilot.com/mcp/ with a PAT or App Token in the Authorization header — no local install needed.
3) Optionally restrict enabled toolsets with --toolsets; only the context toolset is enabled by default.
claude_desktop_config.json
{
"mcpServers": {
"github": {
"command": "docker",
"args": ["run", "-i", "--rm", "-p", "127.0.0.1:8085:8085", "-e", "GITHUB_OAUTH_CALLBACK_PORT=8085", "-e", "GITHUB_PERSONAL_ACCESS_TOKEN", "ghcr.io/github/github-mcp-server"],
"env": {
"GITHUB_PERSONAL_ACCESS_TOKEN": "<YOUR_GITHUB_TOKEN>"
}
}
}
}
Fit and risk
Not for
- Scenarios where you don't want the assistant to have write access to repos (enable only read-only toolsets)
- Environments with strict network isolation for private repos that can't reach the official remote endpoint
Required permissions
- A personal access token (PAT) or OAuth App token; effective scope depends on the token's own permissions
- Enabling toolsets like actions/issues/pull_requests grants write access — request tokens on a least-privilege basis
Risks and side effects
- Write toolsets (creating/merging PRs, triggering workflows) can cause accidental changes if the token is overscoped — try a read-only toolset first
- In hosted mode, credentials travel via the Authorization header — make sure the client-to-api.githubcopilot.com connection is trusted
Troubleshooting
- OAuth callback fails: check whether local port 8085 is in use or blocked by a firewall
- A toolset isn't active: verify the `--toolsets` flag or corresponding env var is spelled correctly
- 401 on the remote endpoint: confirm the PAT/App Token in the Authorization header is valid and not expired
Use cases
Create, comment on, and merge pull requests via natural language
Have an AI assistant diagnose GitHub Actions CI failures
Bulk-manage issue labels and notifications
Query code security scanning results directly from your IDE
Supported clients
Claude DesktopFull support
Claude CodeFull support
VS CodeFull support
CursorFull support
WindsurfFull support
JetBrainsFull support
ZedFull support
AmpFull support