I ship AI-assisted projects to people who cannot read the code, and "the agent says it works" never felt like an honest handover. So I built Mantyl: a free local-first CLI that reads your repo and your coding-agent history, re-runs the checks in a clean Docker sandbox, and compiles a passport your client can independently re-verify with one command.
The passport on the homepage is a real self-run: Mantyl verified its own repository at every milestone, and caught its own undocumented env var doing it.
We go live Tuesday. Before then I would genuinely value eyes on the truth model: every claim is labelled agent said, repository shows, or Mantyl verified, and I want to know what a passport should cover that this one does not. https://www.mantyl.dev/p/495fc5a...