Quorum routes a diff to specialist reviewers — security, performance, correctness, tests, style — in parallel, then a synthesizer consolidates everything into one verdict. Try it live below; it's running the pure-Python engine in WebAssembly.
Each concern gets its own focused reviewer running concurrently — not one model spread thin.
Mock, Claude, OpenAI, or local Ollama — with native structured output and a tolerant fallback.
A detection benchmark scores recall on planted defects and false alarms on clean code.