Just in
Read now

Don’t just measure review bottlenecks. Clear them.

LinearB reviews each pull request as it opens, suggests fixes and routes it to the right reviewer, so your engineers only review what needs a person.
Included in your LinearB subscription at no extra charge

Document the new public API rate limits #4127

OpenSam Ortizwants to merge into main from docs/rate-limits+46 −12 · 3 files
SOSam Ortizopened this pull request · description by LinearB AI
Purpose

Documents the new per-key rate limits for the public API, including the burst allowance, and updates the sample config to match.

Main changes
  • Adds a rate limits section to api/README.md
  • Updates the example limits in config/api.sample.yaml
  • Adds a 429 response example to the errors page
  • Fixes two broken links in the authentication guide
LinearB AIbotreviewed 3 files
Looks good to me
No issues found. The new limits match the values in api/limits.ts, and the sample config uses the same key names as the docs.

LinearB AI approved these changes

gitStream found only docs and config changes, so no reviewer is needed

Approved by LinearB AINo issues found in a docs and config change, so gitStream approved it.

Trusted by the teams transforming software delivery

  • SurveyMonkey
  • Peloton
  • Equinix
  • Norwegian Cruise Line
  • Vizio
  • Gainwell
  • BigID
  • Expedia
  • Syngenta
  • Ingka
  • Silicon Labs
  • Astronomer
  • Mitratech
  • Solace

AI moved the bottleneck from writing code to reviewing it

AI cuts the time it takes to write a change, but every change still waits for a person to review it.
4.6x
longer wait before review for AI-generated pull requests
32.7%
of AI-generated pull requests are accepted, against 84.4% of human-written ones
LinearB benchmarks from 8.1 million+ pull requests
Cycle time breakdownAI-generated PRs against the rest · last 90 days
AI-generated PRs3d 4h
Other PRs2d 17h
CodingPickupReviewDeploy

Sample data

Sort the review queue before anyone opens it

LinearB AI reviews each PR separately from the tool that wrote it. gitStream then approves the clean ones, sends fixes back to the author and routes the rest to the right reviewers.
GitHubGitLabBitbucket
Pull requestsacme/checkout
6 open, waiting for review2 for a reviewer2 with the author2 approved
Update rate limit docs for the public API
Approved
Add CSV export to billing invoices
1 fix suggested
Refresh sessions in the auth middleware
Review requestedsecurity-review
SEC
AG
Move checkout pricing to the pricing service
Review requested45 min review
MBKL
Batch price lookups in the cart
2 fixes suggested
Bump eslint to 9.12
Approved

Set the rules your reviews follow

Write your standards in a REVIEW_RULES.md file or in LinearB settings, for the whole organization or one repository, and LinearB AI follows them in every review.
payments-api/REVIEW_RULES.mdThis repository
  1. 1- Focus on security and performance issues
  2. 2- Flag any hardcoded credentials
  3. 3- Require error handling for all async operations
4Issues limit 105Approve on a clean review Off

Select a rule or setting to see what it changed

LinearB AIbotreviewed 3 files
Securityconfig/stripe.ts:7

A live Stripe key is hardcoded in the config file.

Anyone with read access to the repository can use the key, and rotating it means another deploy.

- const key = "sk_live_51H…";+ const key = process.env.STRIPE_KEY;
Commit suggestion
Performance

One query per invoice in a loop billing/invoices.ts:58

Bug

refund() has no error handling payments/refund.ts:22

3 issuesCommented, not approved

Customers

How engineering leaders cut the time PRs wait for review

magic quadrant mini
Gartner® Magic Quadrant™
LinearB is a Leader
Kraken

Cut review time by more than half in four months

Size and complexity accounted for maybe 5 to 10% of the slowdown. 70 to 80% came from time zone differences, team silos, and reviewer availability.
Nik SudanEngineering Operations Lead, KrakenRead how LinearB shows review load
Epiq

Rebuilt code review for the AI era

Our cycle time wasn't actually improving with AI, because we just moved the bottleneck. It used to be in coding. Now it's in review.
Trevor BestSenior Director of Product Development, EpiqRead the Epiq story
Yum! Brands

Automated over 35% of pull requests

Using the insights and the workflow automations we've set up helps us see where reviews might be stuck, route things smarter, and just get code merged more smoothly between the different groups.
Alex L.Senior Engineering Manager, Yum!Read the Yum! story

Natively secure, obsessively compliant

Independently audited, with evidence of our controls available on request. LinearB participates in the EU-U.S. Data Privacy Framework.
AICPA SOC 1 Type II
AICPA SOC 2 Type II
GDPR: General Data Protection Regulation
ISO 27001 Certified

Automate the rest of your PR workflow

gitStream runs the automations behind the review queue, and you can write your own in YAML.
See workflow automations

AI code review FAQs

GitHub, including GitHub Enterprise Server, GitLab, both SaaS and self-managed, and Bitbucket Cloud. On the Essentials plan, it runs on GitHub only. On GitHub, developers can also run a review on demand by commenting `/gs review` on a pull request.

LinearB reviews a pull request when it opens and when new code is pushed, and skips drafts and PRs opened by bots. Automatic reviews run on the first three pushes to a PR and skip PRs with more than 3,200 changed lines, roughly the largest 5%. On GitHub, `/gs review` runs a review at any time.

LinearB AI reviews code from every assistant and agent your teams use, separately from the tool that wrote it. It reviews each new push, follows the rules you set for your organization, and hands its results to gitStream, which can approve the PR, add labels or request reviewers. It also pulls in the linked ticket from your project management tool and flags gaps between what was asked for and what was built.

Yes. Add guidelines inline or in a REVIEW_RULES.md file for your organization, a team or a repository. You can also set how many issues a comment shows, ask for comments in another language, and approve PRs when the review finds no issues. Admins can test new rules against a real pull request in the AI Playground, in closed beta.

LinearB AI code review runs on Anthropic's Claude Sonnet 4.5 and works with most mainstream programming languages. LinearB adds its own review rules for SQL, BigQuery, TypeScript, C#, SCSS, HCL and YAML.

AI code reviews and AI PR descriptions are included in every paid LinearB plan, including Essentials.

The AI code review dashboard in LinearB tracks AI review coverage, potential issues identified by category, issues resolved and AI-suggested code lines committed, by team, repository and contributor.

Clear the review queue your AI tools filled