LGTM

You context-switch to a browser tab. You skim a 400-line diff. You leave only an LGTM. You context-switch back and try to remember what you were doing. Code review shouldn't break your flow.

LGTM syncs your PRs from GitHub, compiles an AI briefing for each one (risk assessment, flagged areas, draft inline comments) and lets you review and submit without leaving the terminal.

What a briefing looks like

supabase/supabase#19284: Add billing usage export endpoint
author: billing-dev | opened: 2h ago | CI: pass

What changed:
  • New GET endpoint for billing usage CSV export
  • CSV formatting helper for usage data

Risk: high — missing auth middleware on new billing route
Suggested action: request changes | Files: 3 | Lines: +78/-3

[critical] New route exposes billing data without authentication
           src/routes/index.ts:30

3 draft comments. Run lgtm review to walk through.

It learns your codebase

Run lgtm init on a repo and LGTM reads your team's past review comments to bootstrap conventions: patterns like "we always wrap DB calls in transactions" or "errors must use our custom error type." These conventions are checked on every future PR. Over time, it catches the stuff your team cares about without being told.