Show HN: Moa – cross-model peer review for your local agent CLIs (github.com)

A new tool called Moa has been introduced, enabling cross-model peer review for local agent command-line interfaces (CLIs). This tool efficiently distributes a single prompt across multiple installed agents, facilitating simultaneous responses that are streamed back with proper attribution. Moa supports various modes of interaction: it can ask questions, distill multiple responses into a unified answer, or conduct debates between agents, allowing for a collaborative evaluation of their outputs. This functionality is essential for tasks that may benefit from diverse perspectives, as it highlights where different models agree or disagree on specific queries. The significance of Moa lies in its ability to leverage existing subscriptions without requiring additional API keys, thus making advanced comparative analysis accessible to developers. By utilizing a read-only mode by default, Moa ensures the safety of your environment while enabling deeper insights into model reliability. Its structure, with options for customizing input and agent selection, provides a flexible framework for collaborative AI interactions. Furthermore, the built-in functionality for direct adoption of this tool into coding agents enhances its utility within the AI/ML community, fostering an ecosystem of improved decision-making based on model consensus and debate.