Review inspects merge requests with a weighted 6-category scoring rubric. It posts inline comments on specific diff lines, checks your CI pipeline, resolves its own outdated findings, and produces a risk assessment with confidence level.
Reads MR details, the full diff, and all existing discussion threads. Maps every existing comment before forming new findings.
Evaluates new and changed code across six categories: correctness, product behavior, security, performance/quality, tests, and completeness.
Resolves its own outdated findings. Replies to other reviewers' threads. Never creates duplicate discussions.
Produces a weighted score (0-100). 90+ approved, 75-89 review complete, 50-74 changes requested, below 50 blocked.
Correctness (25%), Product Behavior (20%), Security (15%), Code Quality (15%), Completeness (15%), Tests (10%). Weighted to prioritize what matters most.
Posts comments on specific diff lines with severity levels: must-fix, should-fix, and informational.
Reads CI job logs, retries flaky jobs, and penalizes real pipeline failures in the score.
Tracks all existing threads before posting. Never creates duplicate comments. Resolves its own outdated findings.
Evaluates change risk with high/medium/low confidence. Considers scope, complexity, and test coverage.
Makes data-backed approve/block decisions based on scoring thresholds. No subjective vibes — just numbers.
Review Summary
Overall
87/100
Low risk · High confidence