Ship AI code.
Not AI slop.
Codeceipt runs your AI pull request, verifies it actually did what it claims, and gives your client a public receipt they can check themselves.
Verification by execution — not the agent’s opinion of itself.
“✅ All done!” is not proof.
AI agents are confident and wrong. The agent says it’s finished — but nobody ran it. That gap costs both of you.
From pull request to proof in three steps.
The middle step is the whole point: execution, not self-report.
Paste a PR or repo URL
Drop in a GitHub pull request or a whole repository — or add the Codeceipt Action to your repo.
Codeceipt executes it
We run the declared checks ourselves against the diff and the checked-out tree — never the agent’s summary.
Share the Receipt
Get a public link with a verdict your client can re-run themselves. One URL, independently checkable.
This is what your client receives.
One link. They can re-run it themselves. Every line is backed by a real execution.
The one thing your client reads. Plain language, dominant.
Every claim → real executed evidence. Expandable, not a single stamp.
Anyone re-runs it and gets the same fingerprint. You can’t fake it.
Transparent compute & token cost. A literal receipt.
It works because both sides win.
Hand over an executed receipt instead of “trust me.”
Independent proof removes the back-and-forth on quality.
Verified delivery is a premium signal competitors can’t copy.
Embed a verified badge in every PR and README.
We don’t ask you to trust us.
The engine is Apache-2.0. Anyone — including your client — can inspect exactly how a verdict is produced. We give you something you can verify without us.
Free tells you pass or fail.
Five dollars unlocks the whole receipt — every criterion, the fix prompt for what failed, and a link you can hand a client.
Check any public PR or repo. See if it passes — or fails.
Unlock the full proof you hand a client.