PRMergeSafe installs statistics

PRMergeSafe is an AI code reviewer focused on one question: is this PR safe to merge?

We catch the things that break production — breaking changes, logic errors, data risks, security issues, and dependency conflicts. We ignore style nits, naming preferences, and "consider using X" suggestions.

Every PR gets a verdict: SAFE, LOW, MEDIUM, HIGH, or CRITICAL — backed by a multi-model pipeline that escalates only when the stakes warrant it.

What we check

• Breaking changes — removed exports, renamed APIs, signature mismatches, removed routes • Logic errors — off-by-one, null handling, race conditions, broken control flow • Data risks — destructive migrations, missing transactions, schema drift, unsafe deletes • Security issues — hardcoded secrets, SQL injection, auth bypasses, exposed credentials • Dependency conflicts — major version bumps, removed packages, lockfile drift

What we deliberately ignore

• Style preferences and naming nits • "Consider using X instead" suggestions • Refactoring opinions that aren't bugs • Performance micro-optimizations that don't change behavior

How it works

• Skip filter — docs-only and lockfile-only PRs short-circuit without an LLM call • Triage — fast SAFE / NEEDS_REVIEW decision on most PRs • Deep analysis — full review with file context for risky changes • Verification — adversarial second-pass on HIGH/CRITICAL findings to cut false positives

Each finding includes

• What you changed — plain statement of the diff • Why this could break — specific consequence • Where to look — exact file:line references • What to do — concrete fix suggestion

Getting started

• Install the PRMergeSafe GitHub App on your repos • Open a pull request • Get a verdict comment within 60 seconds • Configure custom rules, protected paths, and risk thresholds in your dashboard

Pricing

Free tier available. Paid plans scale by PR volume — same review quality at every tier.

Description from the GitHub Marketplace listing.