Ask a business or strategy question to Claude, GPT, and Gemini simultaneously and see a summary of where they disagreed.
Run a Devil Advocate critique session where one AI model challenges the conclusions of the others on a decision you are considering.
Compare AI model responses side by side with live streaming text and track approximate token costs per session.
Requires your own API keys for each AI provider, the app is unsigned so macOS will prompt for manual approval on first launch.
Council is a native macOS application that sends the same question to several AI models at once and then has them critique each other's answers anonymously. The goal is to surface disagreements between models and give you a more complete picture than asking just one AI would provide. You type a question and up to three advisors answer it simultaneously in side-by-side panels with live streaming text. Each seat can be set to one of nine AI providers: Claude, GPT, Gemini, DeepSeek, Grok, Mistral, Perplexity, OpenRouter, or a locally running model through Ollama. Each seat also gets a role (Analyst, Practitioner, or Skeptic) to encourage genuinely different perspectives rather than three models saying the same thing. A Devil's Advocate role is available to challenge whatever the others conclude. Once all advisors have answered, each one sees the other answers without labels identifying which AI wrote them, and critiques them. The app then shows a Divergence view highlighting where the models disagreed, a Synthesis view summarizing points of agreement, and a full Peer Review page with all the critiques together. You can also drop in an image for models that support vision. Everything is stored locally. Your API keys for each provider are kept in the macOS Keychain and never written to exports, session files, or logs. You pay each provider directly with your own keys, Council does not sit in the middle. There is no account, no server, and no telemetry. A running cost estimate shows token counts and approximate spend as the session progresses. The app requires macOS 14 or later, is built with SwiftUI, and has no third-party dependencies. It can be downloaded as a pre-built binary or compiled from source in Xcode 16. Because it is not signed with a paid Apple developer certificate, macOS will warn on first launch and you need to approve it manually. The project is MIT-licensed and has 17 stars on GitHub.
← albertofettucini on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.