I built this after realizing that that single-model AI code review currently has obvious blind spots. The idea: chat with a Lead AI about your project, then "convene the Council" — three other models independently review everything in parallel, and the Lead synthesizes findings into structured categories: consensus, majority positions, lone warnings, and dissent.

The surprising insight from building and using it: the value isn't consensus, it's structured disagreement. When Grok catches a temporal data mismatch that three other models missed, or Claude spots an imported function that's never called — that's where the Council is useful.

Each model has genuine strengths: Claude tends toward architecture, Grok goes deep on data flows, ChatGPT catches API/integration issues, Gemini thinks about product gaps.

Stack: FastAPI, HTMX, OpenRouter as a unified API gateway. BYOK via OpenRouter so one key covers all four models — a typical review runs about 25 cents. Free tier gives you 1 review to try it.

Perplexity just launched "Model Council" as part of their $200/month Computer product — same core mechanic. This is the open-source, bring-your-own-key version.

GitHub: https://github.com/scifi-signals/council-of-alignment Live: https://council.stardreamgames.com

Constructive feedback would be much appreciated!