Agent Code Reviewer

Helps teams review AI-generated code and catch overlooked risks automatically

Problem

Programmers rely on tools like Claude Code to generate large chunks of code, but many either skip review or can't fully understand the AI's complex logic, letting potential bugs and security holes slip into the repo. On V2EX, someone posted hundreds of lines of agent-generated mask-column transformation code and asked: should agent code be reviewed or not?

Solution

Build a lightweight CLI tool that hooks into Git or PR workflows and performs a 'second review' of AI-generated code: flag suspicious changes, point out boundary conditions missing tests, and highlight removed validation logic. The output isn't a vague score but a line-by-line issue list with fix suggestions that developers can selectively accept.

Users

Individual developers and small teams using AI coding tools like Claude Code or Copilot, especially startups without dedicated code reviewers.

MVP

Parse git diff and AI-generated commit messages; built-in detection rules for common AI code smells; generate line-numbered issue reports; one-click ignore or create fix tasks; run locally without uploading code.

Revenue

Free tier limited to one repo and monthly review count; Pro subscription unlocks multi-repo support, custom team rules, and CI integration.

Why now

AI-generated code volume is exploding while review lags far behind; big vendors won't touch this 'review assistant' niche, and an indie developer can quickly build a rule library and lightweight UI, winning through sensitivity to code smells.