Show HN: Tracking AI Code with Git AI (usegitai.com)

Git AI 1.0 is a new tool that tracks AI-generated code inside Git repos by combining agent hooks with Git-native history tracking. Instead of counting inserted lines, it records precise “checkpoints” when an agent edits files (PreEdit/PostEdit), saves per-commit Authorship Logs as Git notes, and overlays AI authorship on top of git blame so you can see which lines came from which agent. It runs per-machine, offline, wraps your git binary (no repo hooks or background daemons), supports multiple agents (initially Claude Code, Cursor and GitHub Copilot in VS Code), and aims to preserve AI attributions through rebases, cherry-picks, resets, squashes and copies/pastes. This matters because teams need accurate, durable attribution for auditability, collaboration patterns, and measuring how much AI contributes vs. how often its output is rewritten. Git AI avoids heuristics by requiring agents to explicitly mark their output, then reconstructs and rewrites Authorship Logs when history changes using git blame-style diffs. Performance is lightweight (sub-100ms for common commands, under a second for big rewrites), scales with commit size, and stores metadata in Git notes for SCM compatibility and enterprise rollout. Prompt-level capture exists in logs (not yet surfaced in blame), and the project is community-driven with open calls for integrations and platform support.