NEWThree reviewers, one comment, posted as advisory →

A panel of reviewers,
convened on every PR.

Quorum runs Correctness, Security, and Architecture in parallel through OpenRouter on every pull request — then posts one GitHub review with summary + inline comments. Always advisory, never blocks merge. Configurable per-repo via .ai-review.yml.

Install on GitHub Read the docs⌘D

Trusted by teams atLINEARPERPLEXITYVERCELRUNWAYMERCURYSUPABASE

github.com/linear-rb/api/pull/2841quorum-botcommented just now

needs attn3 reviewers, 4 findings · 00:42 · $0.012run 8f2a · 3 / 3 reviewers ✓

Ccritical

94%Security

Webhook signature verification skipped on retry path

src/billing/webhooks.ts:142

Hhigh

86%Correctness

Race condition between webhook handler and seat refresh

src/billing/sync.ts:73

Hhigh

79%Architecture

Stripe client constructed inside the handler — leaks across handlers

src/billing/index.ts:21

Mmed

68%Correctness

Idempotency key derived from event.id only

src/billing/webhooks.ts:89

How it works

Three reviewers, one verdict, posted inline.

Each reviewer has its own focus list and an optional model override. The aggregator filters by min_confidence, dedupes by (file, line, title), and posts a single review.

Per-repo policy

.ai-review.yml at the repo root. Trigger events, draft skip, min_confidence, max_inline_comments, per-reviewer model overrides.

OpenRouter through-routing

All three reviewers call OpenRouter chat-completions in parallel. Default ling-2.6-1t (free); override per reviewer.

Inngest workflow

Background reviewer fan-out with concurrency 5 per repo and 2 retries. Run state persists in review_runs.

Single GitHub review post

One POST /pulls/n/reviews carries summary + inline comments at lines that map to a diff position. Always 'COMMENT'.

Quota enforcement

review_runs counted per calendar month against your active stripe_subscriptions row. Hard-stop on overage.

Cost ledger

Per-reviewer prompt + completion + total tokens recorded in usage_events. Aggregated cost shown on every run.

The pipeline

From webhook to inline review.

PR opened

github webhook

Diff fetched

patch + context

Correctness

Security

Architecture

Reviewers

3 in parallel

Aggregator

dedupe · rank

Posted

inline + summary

Configurable

One YAML file. Or none — sensible defaults.

Drop .ai-review.yml in your repo to override the dashboard policy. Friendly form view if you'd rather click. Diff against the org default at a glance.

Per-repo agent enable/disable
Skip globs and labels (e.g. don't review docs-only PRs)
Confidence thresholds and max inline comments
Custom prompt instructions per agent

.ai-review.ymlYAML · 21 lines

1review:
2  enabled: true
3  trigger:
4    - opened
5    - synchronize
6 
7  skip:
8    draft_prs: true
9    labels: ['skip-ai-review']
10 
11  rules:
12    min_confidence: 0.75
13    max_inline_comments: 10
14 
15  reviewers:
16    - id: correctness
17      model: inclusionai/ling-2.6-1t:free
18    - id: security
19      model: inclusionai/ling-2.6-1t:free
20    - id: architecture
21      model: inclusionai/ling-2.6-1t:free

Pricing

Pay for the reviews you actually run.

Free

$0forever

Trial, OSS, hobby.

Install on GitHub

10 reviews / mo
1 private repo
Summary + max 3 inline
Basic model preset

Developer

$19/ mo

Solo devs and indie builders.

Upgrade

100 reviews / mo
5 repos
quiet · balanced · paranoid
BYO OpenRouter key

A panel of reviewers,convened on every PR.

Three reviewers, one verdict, posted inline.

From webhook to inline review.

One YAML file. Or none — sensible defaults.

Pay for the reviews you actually run.

A panel of reviewers,
convened on every PR.