Who actually wrote this code?

Git tells you what changed. Entire tells you why, and who. Or what.

Here's the problem. You're running Claude Code, Codex, Gemini CLI, whatever. The agent writes a hundred lines, you tweak ten, commit, push. Six months later someone's debugging that function and git blame says your name. But you didn't write most of it. You don't even remember what prompt produced it. The reasoning, the agent's decisions, the tool calls, all gone.

Entire hooks into your git push and captures the full AI session. Prompts, responses, tool calls, files touched, token usage. Everything gets stored on a hidden branch (entire/checkpoints/v1) so your main history stays clean. Each commit gets a 12-character Checkpoint ID linking back to the session on their dashboard.

The line-level attribution is the bit that matters. Not just "AI helped with this file" but actual percentage breakdowns of which lines were agent-written versus human-written. For audits, for onboarding, for debugging at 2am when you need to understand intent behind code you didn't write. Proper useful.

Thomas Dohmke (ex-GitHub CEO) started this with a $60M seed. Agent-agnostic by design, works with Claude Code, Codex, Gemini CLI, Cursor, the lot. I reckon every team shipping AI-assisted code daily is going to need something like this eventually.