Git

The Git server exposes repository operations — status, diff, log, branch, add, commit — as structured tools instead of shell commands, which means an agent can inspect and act on a repository’s history without you having to grant it raw terminal access. Point it at a repository with --repository /path/to/repo and it operates against that working tree for the life of the session.

It’s a natural companion to the Filesystem server: Filesystem handles reading and writing file contents, Git handles understanding and recording the history around those changes. Together they cover most of what a coding agent needs for local, non-destructive repository work — reviewing a diff before proposing a change, drafting a commit message that actually reflects what moved, or walking through recent log entries to understand why a file looks the way it does.

No environment variables or API keys are required; the only configuration is which repository path to operate against.