From the moment a PR opens, PR Quorum reads the diff like a disciplined review team: specialists fan out, noisy findings get filtered, and GitHub receives one focused review that maintainers can act on.
01 · Anatomy
The review pipeline
Each PR triggers a fan-out across specialist reviewers. They return typed findings, the aggregator removes repeats, and only the strongest comments reach GitHub.
GitHub webhookpull_request.opened | synchronize
Inngest functionconcurrency 5/repo · 2 retries
Reviewers (parallel)OpenRouter chat-completions
AggregatorZod-validated · dedupe · severity sort
Single review postsummary body + inline at diff position
PR opened
acme-payments/api · #2841
0.0s
Webhook received
pull_request.synchronize
0.2s
Diff fetched
46 files · +1,284 / −188
0.8s
Reviewers · 3 in parallel
~22s
Correctnessdeepseek-v4-flashfound 3
Securitydeepseek-v4-flashfound 1
Architecturedeepseek-v4-flashclean
Aggregator
filter < min_confidence · dedupe · sort by severity
< 0.1s
Posted to GitHub
POST /pulls/n/reviews · event=COMMENT
0.3s
total23.9s
02 · Why teams ship faster
Built for signal, not volume
DEDUPE
One signal per defect
Multiple reviewers can spot the same issue. PR Quorum merges duplicates, ranks by severity and confidence, then posts only the comments worth reading.
min_confidence: 0.75
max_inline_comments: 10
POLICY
Repo-specific review taste
.ai-review.yml lets every repo decide what strict means: hotter auth paths, quieter generated files, draft skips, and per-reviewer model overrides.