AI Agents Hub
CodeRabbit
AI PR reviewer — line-by-line comments on every PR within minutes, free for OSS.
The default AI PR reviewer on GitHub — line-by-line comments, walkthrough of changes, and a chat to ask follow-ups. Free for OSS, $12/dev/mo otherwise.
What it is
CodeRabbit is an AI code reviewer that lives inside pull requests on GitHub, GitLab, Bitbucket, and Azure DevOps. Within ~2 minutes of opening a PR, it posts a top-level walkthrough plus a stream of inline line-by-line comments with concrete fix suggestions (often code-diff level). It is aware of your formatter (Ruff, Mypy, ESLint, Prettier, Clippy, etc.), your test runner, and your CI failures, and it can produce both stylistic and logical feedback. Pricing tiers: free for public/open-source repos; Pro at $12/dev/mo for private repos with full review features; Enterprise tier adds SSO, audit logs, custom models, RBAC, and on-prem deployment.
Best for
- Pull-request reviews on private repos with concrete fix suggestions
- Style + lint enforcement at PR time (Ruff, Mypy, ESLint, Prettier-aware)
- Walking a new contributor through a complex PR via walkthrough + chat
- Free OSS repo reviews with no per-dev cost
Pros
- Free for public/open-source repos — no per-dev cost on OSS
- ~2-minute review SLA per PR — fast turnaround
- Walkthrough + chat mode for follow-up questions on the same PR
- Tooling-aware: integrates with your formatter and linter configs
Cons
- PR review isn't a substitute for end-to-end coding agents (Cursor / Windsurf)
- Free OSS tier is review-only; Pro $12/dev/mo is per-developer (gets expensive on large teams)
- Cloud-only (Enterprise tier supports on-prem)
Quick start
- Go to coderabbit.ai → "Sign in with GitHub" (or GitLab / Bitbucket / Azure DevOps).
- Install the GitHub App on your org or specific repos. Free on public repos; Pro $12/dev/mo on private.
- Open any PR → CodeRabbit posts a review within 1–3 minutes with inline comments + summary.
- Click "Chat" on the PR → talk to CodeRabbit to ask for changes, fixes, or clarifications.
- Add a `.coderabbit.yaml` to customize review focus (style guide, auto-suggestions, walkthrough level).
- Enterprise tier: SSO, audit logs, custom models, RBAC, and on-prem install for regulated teams.
Sample input / output
Repo: acme/api-server (private, Python / FastAPI)
PR #1842: "Add /v1/billing endpoint"
CodeRabbit config (.coderabbit.yaml):
reviews:
profile: detailed
auto_suggestions: true
walkthrough_level: high
tools:
ruff: enable
mypy: enable
PR diff (+220 lines): adds billing.py, tests, alembic migration.CodeRabbit review (2m 11s after PR opened):
Summary comment on PR:
"Adds /v1/billing with 3 distinct handlers, 6 new DB columns, and 8 test cases. Suggests 4 refactors before merge: (a) split the 142-line billing_service.py into 3 single-responsibility modules, (b) replace manual Decimal arithmetic with `decimal.Decimal.quantize`, (c) extract retry policy into the stripe abstraction, (d) tighten test isolation."
Inline comments (12 total, 3 actionable):
- billing.py:84 — "Use `Decimal.quantize(Decimal('0.01'))` instead of rounding manually"
- billing.py:131 — "Consider `stripe.Transfer.create(..., idempotency_key=...)` to prevent double-charges"
- tests/test_billing.py:62 — "This test depends on real Stripe — use `stripe-mock` for CI"
Walkthrough mode: "Path A → Customer lookup → Decimal math → Stripe transfer → Response"
Chat: "What's the failure mode if Stripe returns 502? Re-route to retry queue? Add a circuit breaker?"
After applying suggestions: 4 file changes, ruff + mypy clean, all 8 new tests pass.Benchmarks
Pricing
Free for public repos; Pro $12/dev/mo for private repos (per-developer billing); Enterprise custom with SSO/audit/on-prem.