How is MicroReview different from CodeRabbit?

MicroReview adds a 0-100 risk score per PR, merge blocking on critical findings, a dedicated GitHub Checks API integration with inline annotations, and hardcoded secret detection across 13 patterns. It also starts free and scales to $19/mo per repo vs CodeRabbit's higher per-seat pricing.

Does MicroReview work with private repositories?

Yes. MicroReview works with both public and private repositories on GitHub and GitLab. The GitHub App requests only the minimum permissions needed: read access to code for PR diffs and write access to post review comments and check runs.

What programming languages does MicroReview support?

AI bug detection and secret scanning work on any language. Deep static analysis rules including naming conventions, Spring Boot patterns, and validation checks are available for Java, TypeScript, and Python today with more language-specific rule sets coming soon.

Is my source code sent to AI?

Only the files your pull request touches are sent to the AI — the changed lines plus the surrounding file and the local files it imports, which is what makes the review accurate instead of guessing at missing context. Your wider repository is never uploaded. Code is sent over HTTPS, is not used to train models, and is retained by the AI provider for at most 30 days for abuse monitoring before deletion.

How much does MicroReview cost?

MicroReview starts free with 2 repos and 50 reviews per month. The Pro plan is $19/mo per repo with unlimited reviews and AI bug detection. The Team plan is $15/mo per repo for 5+ repos with analytics, SSO, and config inheritance.

Changelog — MicroReview

Built in the open by one engineer. Dates are real ship dates, and the accuracy items come straight from tester feedback.

July 25, 2026Site

One design system across site and app

The marketing site and the logged-in dashboard now share the same palette and surfaces, and every GitHub link points at the Marketplace listing. Also fixed a stale claim: the AI reads full changed files plus their local imports (that's what makes it accurate), never the wider repo.

July 17, 2026Accuracy

Why diff-only review produces false positives — write-up

A full engineering post on the false-positive class an early tester reported, the root cause, and the repo-context architecture that fixed it. Read it on the blog.

July 13, 2026Dashboard

Turn repos off, keep their history

You can now disable reviews on a repo from its page — billing shrinks accordingly and all review history is kept for when you re-enable. Enabling a repo that was reviewed via a Marketplace install now adopts its past reviews, so history appears immediately. Plus: configuration docs at /docs/config, and every PR summary now links back to your dashboard.

July 11, 2026Dashboard

Dashboard redesign shipped

New app shell with sidebar navigation, a repos overview with risk sparklines, real-data analytics (risk over time, top rules, auto-approve rate), redesigned PR detail and repo pages, and a visual config editor with live YAML preview. Navigation is near-instant: the shell renders once and content streams in.

July 11, 2026Billing

Fair free-tier enforcement + upgrade flow

The free tier's 50 reviews/month is now enforced per account (not per repo), with a single courtesy email the first time you hit it. Upgrading to Pro is one click from the sidebar or Settings. The repos table shows your real quota instead of a misleading ∞.

July 9, 2026Accuracy

Full-file context: the false-positive fix

The AI reviewer now reads each changed file's complete contents and the local files it imports before flagging anything — so it verifies imports, type definitions and component sources instead of guessing from the diff. This removed the "X is not imported" class of false positive entirely.

July 8, 2026Platform

Marketplace installs, first-class

Repos installed from the GitHub Marketplace are tracked from the moment of install — reviews run and are recorded even before you ever sign in to the web dashboard.

July 7, 2026Accuracy

Upgraded review model + quieter approvals

Reviews now run on a stronger frontier model with a severity rubric that reserves "critical" for genuinely severe issues. Fixed duplicate "looks good" approvals on every push — one approval, maintained across commits.

June 16, 2026Accuracy

Severity calibration

Added a conservative severity rubric to the AI reviewer so critical findings stay rare and meaningful — a typical PR has zero or one, and the merge gate only fires when it should.

June 14, 2026Platform

GitHub Marketplace listing approved

MicroReview is installable directly from the GitHub Marketplace.

What's shipping