Agent Blame. Track AI-code in your repos.

git blame for the AI-era

git blame shows you which human committed a line so you know who to talk to about any hunk of code. But when most code (>90%) is written by the Agents who do you talk to when you don't understand something? Develoeprs move so fast with agetns that their intent, architectual decisions and learnings during implimentation are not being documented.

Git AI fixes this. It's an open-source Git extension that gives you git ai blame — the same familiar interface, but with full AI attribution on every line. Keep track of which model, agent, and prompt generated each line of code.

Git AI is a native Git extension. AI attributions are stored in Git Notes — a first-class Git feature — not in proprietary databases, metadata services, or vendor-locked platforms. This means:

• Attributions live in your repo. They travel with your code through clones, forks, and mirrors.
• No vendor lock-in. The attribution format is an open standard anyone can read and write.
• Works offline. No network calls, no API keys, no login required.
• Survives everything Git throws at it. Rebases, squashes, cherry-picks, amends, merges, resets — Git AI rewrites attributions automatically so they're never lost.

Zero workflow changes

Install the extension once. There's nothing to configure per-repo and no hooks to manage. Developers keep working exactly as they do today. When a developer commits, Git AI automatically records which lines were written by AI and which were written by the human:

That's it. No extra commands, no manual tagging, no disruption. The attribution happens in the background using agent hooks that the supported coding agents already understand.

Works with every major coding agent

Git AI integrates with all the popular coding agents. Each agent reports exactly which lines it generated — no scanning, no detection, no guessing.

Cursor, Claude Code, OpenAI Codex, GitHub Copilot, Gemini CLI, OpenCode, Continue, Droid, Junie, Rovo Dev, Amp, Windsurf

How it works

1. Agents call git ai checkpoint as they code — marking the code they generate explicitly. This is implemented as a mix of hooks, first-party integrations, and plugins — no workflow changes required.
2. AI attributions are saved in Git Notes on commit — When you commit, all checkpoints are condensed into an Authorship Log and attached to the commit as a Git Note.
3. Attributions survive history rewrites — Git AI automatically rewrites Authorship Logs after rebases, squashes, cherry-picks, amends, merges, and resets, so AI attribution is never lost.

Track AI code through the entire SDLC

The Coding Agent companies stop at "lines generated." But what matters is how much AI code gets committed, makes it through code review, ships to production, and holds up over time.

Git AI is the only tool that can track AI-authored code through every stage: Generated → Committed → PR Review → Merged → Production → Durable.

• AI Effectiveness — for every line we generate, how much makes it to production (average 20:1)
• Authorship tracking — know which agent generated every line and who approved it
• Incident correlation — connect production issues back to AI-generated code
• Audit & compliance — prove exactly which code was AI-generated, by which agent, and who approved it
• Code durability — measure how long AI code survives before being removed or rewritten