CS student at HS RheinMain in Wiesbaden, Germany.
I'm building MergeMonkey, a code reviewer for pull requests. Not purely static analysis, not just an LLM eyeballing the diff.
The output looks like: "userId comes in from req.params.id at line 38, passes through formatUser() with no sanitization, reaches db.query at line 42" - rather than "this looks like SQL injection."
How it works:
- Two-pass review: fast lint pass (catches obvious stuff) + deep agent investigation (traces data flows, finds security issues)
- Bring your own model via OpenRouter, no lock-in to one provider
- Supports GitHub and GitLab (including self-hosted)
What I care about: security, code quality, flexibility and making code review less painful for maintainers drowning in PRs.