Git AI

CLI Reference

Complete reference for all Git AI commands including blame, stats, checkpoint, and install-hooks. Learn the full command-line interface for tracking AI code authorship.

When you invoke the Git AI binary as git it will proxy the arguments to git and pipe the output and exist code back to you.

To run Git AI specific commands you invoke the same binary as git-ai. The symlinks and $PATH export the install scripts setup sets all this up on your machine.

git-ai <command> [options]

User Commands

blame

Enhanced version of git blame that shows AI authorship attribution alongside traditional git blame.

git-ai blame <file>

Arguments:

  • <file> - Path to the file to blame (required)

Options: Mostly API Compatible, supports same options as git blame.

stats

Show AI authorship statistics for a commit. Displays how much code was written by humans vs AI.

# Show stats for current HEAD
git-ai stats

# Show stats for specific commit
git-ai stats <commit-sha>

# Output in JSON format
git-ai stats --json
git-ai stats <commit-sha> --json

Options:

  • <commit-sha> - Optional commit SHA (defaults to HEAD)
  • --json - Output statistics in JSON format
install-hooks

Automatically configure Claude Code, Cursor and GitHub Copilot to send authorship information to the git-ai binary

git-ai install-hooks

Plumbing Commands (for AI Agents)

These commands are primarily used by AI agents (Claude Code, Cursor, GitHub Copilot) for authorship tracking. End users typically don't need to use these directly.

checkpoint

Checkpoint working changes and attribute them to an AI agent or human author.

# Basic checkpoint with preset
git-ai checkpoint <preset>

# Show the current working log
git-ai checkpoint --show-working-log

# Reset the working log
git-ai checkpoint --reset

Presets:

  • claude - Claude AI integration
  • cursor - Cursor editor integration
  • github-copilot - GitHub Copilot integration
  • mock_ai - Mock AI for testing

Options:

  • --show-working-log - Display current working log without making changes
  • --reset - Clear the working log
squash-authorship

Generate authorship information from squashed commits. Used when commits are squashed to reconstruct authorship metadata.

git-ai squash-authorship <branch> <new-sha> <old-sha>
git-ai squash-authorship <branch> <new-sha> <old-sha> --dry-run

Arguments:

  • <branch> - Branch name (required)
  • <new-sha> - New commit SHA after squash (required)
  • <old-sha> - Old commit SHA before squash (required)

Options:

  • --dry-run - Show what would be done without making changes
git-path

Print the path to the underlying git executable. Used by integrations to locate the git binary.

git-ai git-path

Version

Show the version of git-ai:

git-ai version
git-ai --version
git-ai -v

How Git AI Works

Learn how Git AI tracks AI code authorship through checkpoints, authorship logs, and git notes. Understand the technical design behind accurate AI attribution through the entire development lifecycle.

Limitations

Supported AI agents, git operations, and known limitations of Git AI. See which features work with Cursor, Claude Code, GitHub Copilot and which workflows are supported.

On this page

User Commandsblamestatsinstall-hooksPlumbing Commands (for AI Agents)checkpointsquash-authorshipgit-pathVersion