AI Code Review for TypeScript Projects
The compiler proves your types line up. It does not tell you the promise you forgot to await will silently swallow an error, or that a React effect re-subscribes on every render. That is the gap a reading-based review fills.
What the review catches in TypeScript
- any and unchecked casts that erase the guarantees the rest of the file relies on.
- Unawaited promises, floating async calls and unhandled rejections.
- Narrowing that looks safe but does not hold at runtime, such as trusting a parsed JSON shape.
- React issues: missing effect cleanup, unstable dependency arrays, state updates after unmount.
- Error paths that return a success shape, and catch blocks that discard the cause.
- Client code trusting data that has not been validated at the boundary.
Works with your existing tooling
ESLint and tsc stay exactly where they are. They enforce rules; this review reads intent. The two rarely overlap, and the review is deliberately quiet about anything a formatter would already fix.
Fixes you can apply
Each finding comes with the current code and a proposed replacement. On GitHub pull requests the change can be committed to the branch after the file is verified, so a suggestion built from an older version never overwrites newer work.
Frequently asked questions
Does it understand React and Node together?
Yes. Reviews read the change in the context of the surrounding file, so client and server code in the same pull request are both analysed.
Will it fight my lint config?
Stylistic nitpicks can be switched off so the review reports correctness and security issues only.
What about plain JavaScript?
Fully supported. Type annotations help the review, but are not required.
Free plan available. No credit card required.