Get Started

Quick Start

Mac, Linux, Windows (WSL)

curl -sSL https://usegitai.com/install.sh | bash

Windows (non-WSL)

powershell -NoProfile -ExecutionPolicy Bypass -Command "irm http://usegitai.com/install.ps1 | iex"

🎊 That's it! No per-repo setup. Once installed Git AI will work OOTB with any of these Supported Agents:

Next step: Just code and commit!

Once installed, all your Coding Agents will call Git AI in the background and mark the code they generate AI-authored.

After you commit, git-ai adds a git note to track which lines were AI-authored and scores each commit:

Installing the Stats Bot (early access)

Aggregate git-ai data at the PR, developer, Repository and Organization levels:

• AI authorship breakdown for every Pull Request
• Measure % of code that is AI generated through the entire SDLC
• Compare accepted-rate for code written by each Agent + Model.
• AI-Code Halflife (how durable is the AI code)

Get early access by chatting with the maintainers

Prompt Storage

By default Git AI stores prompt data locally only. To include prompts in git notes (authorship logs), set prompt_storage to notes:

git-ai config set prompt_storage notes

When using notes mode, you can exclude specific repositories from having prompt data included:

git-ai config set --add exclude_prompts_in_repositories https://github.com/private-org/*
git-ai config set --add exclude_prompts_in_repositories /path/to/private/repo

or to exclude all repositories:

git-ai config set --add exclude_prompts_in_repositories "*"

Goals of git-ai project

🤖 Track AI code in a Multi-Agent world. Because developers get to choose their tools, engineering teams need a vendor agnostic way to track AI impact in their repos.

🎯 Accurate attribution from Laptop → Pull Request → Merged. Claude Code, Cursor and Copilot cannot track code after generation—Git AI follows it through the entire workflow.

🔄 Support real-world git workflows by making sure AI-Authorship annotations survive a merge --squash, rebase, reset, cherry-pick etc.

🔗 Maintain link between prompts and code - there is valuable context and requirements in team prompts—preserve them alongside code.

🚀 Git-native + Fast - git-ai is built on git plumbing commands. Negligible impact even in large repos (<100ms). Tested in Chromium.

Documentation

• How Git AI Works and its Limitations ▶️ Video 🗺️ Diagram
• Git AI Commands
• Configuring Git AI for the enterprise

Agent Support

git-ai automatically sets up all supported agent hooks using the git-ai install-hooks command

| your agent here | | |

Building a Coding Agent? Add support for Git AI by following this guide