Find the bugs your coding agent leaves behind
Safe Seer is a static analyzer for security and correctness defects: injection, broken access control, leaked secrets, unverified webhooks, null dereferences, resource leaks and more. It runs inside Claude Code, Codex and Cursor, and your agent triages and fixes what it finds.
Install
In your project (Node 20 or later):
npx safeseer init
It connects your coding agent, scans the whole repository once, and re-scans incrementally after every commit.
How it works
- The analysis runs on your machine. Your code is not uploaded.
- Each defect comes with its data flow and a stable id.
- Your agent fixes it, or records why it is a false positive or deferred; the next scan confirms the fix.