git-ai

Why we included this project

If your team leans on AI coding agents, you've probably wondered which lines in the codebase came from a model and which were written by hand. git-ai answers that by tying each line to the agent, model, and prompt that produced it, so a single command tells you exactly what an AI wrote and why. It drops into your normal workflow with no per-repo setup or git hooks, and it also reports aggregate numbers like the share of AI code, how much of that gets accepted versus rewritten, and a breakdown by tool and model. That makes it useful for code review, auditing AI-assisted work, and spotting which coding tools actually hold up over time. Attribution lives locally in git notes with prompts redacted and out of the repository, and it survives rebases and merges, so it behaves more like a durable analytics layer than a one-off reporting script.