Code review that knows how apps break
Cody is the reviewer Chirai runs on every app it builds, now for your own code. It reads a repository and points to the exact file and line of each problem, with why it matters and the fix. It skips style nits.
What it finds
Live Stripe, AWS, OpenAI or GitHub keys in code, a committed .env, a Supabase service_role key, server secrets in client code.
Supabase tables with no row-level security, and policies that let anyone, or any signed-in user, change rows.
useSearchParams without Suspense, metadata in client components, hooks in server pages, imports of files that don’t exist.
Links and redirects to pages or screens that don’t exist, in Next.js and expo-router.
Text outside <Text> in React Native: it shows in the web preview and crashes on a real phone.
Wrong results, missing ownership checks, injection, race conditions and data loss, each on a real line.
Ways to use it
Sign in and paste a GitHub repository (a token for a private one is used once and never stored).
No limits. Add --ai with your own Anthropic or OpenAI-compatible key for the AI reviewer.
npx chirai-cody . --ai
Cody comments on the changed lines and gives the change a score; errors can fail the check.
npx chirai-cody . --diff origin/main --github
Works with JavaScript and TypeScript: Next.js (app and pages router), React, Expo / React Native, Node and Supabase. Reports in the terminal, Markdown, JSON, HTML or SARIF for GitHub code scanning.