AI code review
Find the problems before the reviewer does.
Paste a file. Get every bug, security risk and performance problem found in it — each one explained, located, and paired with a fix.
Then a score out of 100 across seven categories, with the arithmetic shown. The AI proposes the category scores; the backend validates them and calculates the total.
What it finds
Six kinds of problem, one explanation each.
Null dereferences, off-by-one mistakes, unhandled edge cases, and control flow that does not do what it appears to.
Injection risks, unsafe deserialization, hard-coded secrets, and input that reaches sensitive operations unvalidated.
Avoidable algorithmic cost, repeated work, N+1 queries, and resources such as files or connections that are never closed.
Unclear naming, oversized functions, duplication, and departures from the accepted idioms of the language.
Tight coupling, mixed responsibilities, and structure that will make the next change harder than it needs to be.
Seven categories out of 100, each with feedback, strengths and improvements, plus the arithmetic behind the total.
How it works
Four steps, no sign-up.
- 01 Pick a language
Choose from the languages the backend accepts, and optionally name the file.
- 02 Paste your code
Add optional instructions to steer what the review emphasises.
- 03 The backend reviews it
Django calls the AI provider, then validates every category score against the marking scheme.
- 04 Read the result
Issues with explanations and fixes, plus a score breakdown showing exactly how the total was reached.
Your code reaches the AI provider through the backend only. The API key never reaches the browser, and neither your code nor any suggested fix is ever executed.
Marking scheme
The score is the sum of its parts.
Published by the API, not hard-coded in this page — so the criteria you read here are exactly the ones your code is marked against.