ReproPatch Studio

ReproPatch Studio is an AI repair studio for developers that turns a bug report into a verified code patch.

Instead of asking AI to simply “suggest a fix,” ReproPatch runs a full engineering repair loop: it creates a safe workspace copy of the codebase, uses Codex as a coding repair agent, generates a regression test that reproduces the bug, patches the source code, runs the test suite again, and then presents a PR-ready diff and summary.

The interface is built like a mission-control dashboard so the whole process is visible and inspectable: bug report, agent timeline, failing test, generated patch, passing tests, changed files, risk notes, and final ready-to-review state.

For the hackathon version, ReproPatch runs locally on a controlled TypeScript demo repo with three deterministic bugs: ROI zero equity, cart zero quantity, and todo active-filter mismatch. It also includes golden fallback runs so the demo remains reliable even if live Codex/auth/Wi-Fi fails during judging.

ReproPatch is for developers, open-source maintainers, and engineering teams who want AI coding agents that do more than generate plausible code — they prove their work.