GuidesCode ReviewReview policy

Review Policy

What this page covers

How Mesrai’s review output interacts with your branch protection rules, and when to dial up enforcement versus stay advisory.

Three policy modes

  • Advisory (default) — Mesrai posts non-blocking comments. Your team decides what to act on.
  • Request changes — opt in. Mesrai marks the review as changes requested when it finds an issue above your severity threshold. Branch protection then blocks merge.
  • Auto-approve — opt in. Mesrai approves the PR when it finds nothing above threshold. Useful when an approval from the bot satisfies a required-reviewer count.

Enable the enforcement modes only once your team has aligned on the severity threshold and has CI you trust. Premature blocking creates friction; conservative thresholds buy trust.

When Request changes makes sense

  • Security-critical paths where a critical finding must never merge
  • Performance hot spots where regressions are expensive to roll back
  • Mature repositories with a settled set of review rules

When Auto-approve makes sense

  • Small, low-risk diffs (docs, tests, dependency bumps)
  • Teams with strong CI + high test coverage
  • Org policies that require a bot approval as one of several required approvals

Humans still own the call

Mesrai accelerates the feedback loop — it does not replace human judgment. Final merge decisions belong to your team. The right blocking policy is the one your team will defend on a Friday afternoon.