# ChainReview

- **Event:** [Built with Opus 4.6: a Claude Code hackathon](https://cerebralvalley.ai/e/claude-code-hackathon)
- **When:** Feb 10 at 12:00 PM – Feb 17 at 10:00 AM (EST)
- **Where:** Location TBA
- **Team:** [Charann Srinivas](https://cerebralvalley.ai/u/charann)
- **GitHub:** https://github.com/charannyk06/ChainReview
- **Demo video:** https://youtu.be/MVYIiM0yKTc?si=dfm4F4klorZ4F_Fi
- **Gallery:** https://cerebralvalley.ai/e/claude-code-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/claude-code-hackathon/hackathon/gallery/36

Vibe coding makes it ridiculously easy to ship. You generate features in minutes, hit run, and everything looks fine—until the first real edge case, the first PR review, the first CI failure. Then the "almost right" code shows up as noise: false positives, shallow findings, and hours lost in back-and-forth.
We built ChainReview to turn that chaos into a closed loop. ChainReview is a multi-agent AI code reviewer where five specialists—Architecture, Security, Bugs, Explainer, and a strict Validator—work in parallel. Every agent is grounded in deterministic tooling via CRP (ChainReview Protocol), an open MCP-compatible schema with 20+ tools: file reads, import graphs, call graphs, Semgrep scans, and ripgrep pattern matching. Every finding is pinned to real evidence—file paths, line numbers, and confidence scores.

The twist is the Validator Agent. It challenges every finding, re-reads the code, checks context, and kills the noise. What survives is real, actionable, and ready to fix.

---

Markdown version of https://cerebralvalley.ai/e/claude-code-hackathon/hackathon/gallery/36. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt