OhMyBugAnswers in
plain English.
How reviews work, what they cost, and how we handle your code.
01 / The review
What does OhMyBug actually do?
It reviews a code change before you merge it. Independent review passes look for bugs. Your coding agent checks each finding against the repository, then you approve the verdicts.
Why not ask my coding agent again?
You should. OhMyBug gives the change a separate review context and a skeptic who tries to disprove possible bugs. It is a second check, not a replacement for your agent or your own review.
When should I run it?
After your review and CI, on the exact diff you plan to merge. Reviewing an older diff does not cover today’s changes.
Can I use Claude Code or Codex?
Yes. Pick your agent on the landing page for the install command. Sign in with GitHub once; you do not need a model-provider API key.
How long does a review take?
It depends on the size of the change, the available context, and the review mode. Your agent shows progress and, when we have enough data, a recent median. A deep review takes longer.
02 / The price
What do I pay for?
A review costs $10 when at least one medium, high, or critical bug is confirmed. One bug or ten costs the same. Clean reviews, rejected or unclear findings, and minor-only findings cost $0. Your first bug-finding review is free.
When does billing happen?
When your agent submits the verdicts—not when the review starts. New reviews stop when your credit runs out, but reviews already started can still be billed. See the terms for packs and subscriptions.
03 / Your code
What leaves my machine?
Your diff and the context files you approve. Reviewers may ask for more. For a public repository—or one connected through the read-only GitHub App—the service can fetch available files. A full-repository review is optional and starts only after you approve it.
Do you keep my code?
We do not keep a permanent copy of your repository. Submitted files leave the database after the sandbox receives them, though a retry may store them again briefly. Run transcripts stay for 7 days after a successful review or 30 days after a failed one. Reports and findings can quote code and remain in your review history. See the privacy policy.
Is my code used for model training?
No. We do not train models on your submissions. Model providers process them under their API terms; the providers and their commitments are in our privacy policy.
Does a clean review mean no bugs?
No. Reviewers can miss bugs or be wrong. Keep your tests and your own review. See what the review does not guarantee.
Want the mechanics?
See how it works →