OmniCouncil – Zero-API-cost local MoA desktop app for Mac (github.com)

OmniCouncil has launched a local-first Mixture-of-Agents (MoA) desktop app for macOS, allowing users to query multiple AI models—such as Claude, Gemini, and Codex—simultaneously. The app operates in a zero-API-cost environment by using each model's CLI through the user's existing subscriptions, eliminating the need for separate API keys or per-token billing. Its dual modes, Judge and Co-work, facilitate independent model answers that are evaluated for consensus, enhancing accuracy by revealing discrepancies among different AIs. This app holds significant implications for the AI/ML community as it advocates a collaborative approach to leveraging AI models, potentially leading to higher accuracy and error detection compared to using a single model. In Judge Mode, three models respond independently and are assessed by a Judge model, which provides a consensus score and initiates a second review if needed. Co-work Mode promotes interactive discussions among the models to refine answers. The app's performance was validated in pilot tests showing improved accuracy when multiple models participated, demonstrating the potential to tackle complex queries more effectively.