ministry-of-experts

Use the /moa command followed by your question. The lead model sends the identical question to expert models like DeepSeek V4, GLM 5.2, and GPT 5.5, then synthesizes their responses into one ranked answer.

Mixture-of-Agents routes one question to several expert models and has a lead model synthesize their responses. This catches individual model biases and forgotten points, producing a ranked answer with attribution for each claim.

Yes, it works directly in Hermes with no external UI required. Invoke it by typing /moa followed by your question, and the lead model coordinates the expert panel within the chat.

The lead model reads the three raw expert responses and compresses them into a single synthesis. You consume one consolidated answer instead of reading three full model dumps, reducing token spend.

Avoid it for simple factual lookups or quick tasks where one model suffices, since querying multiple models costs more tokens and time. Reserve it for complex, high-stakes questions where answer quality justifies the overhead.