Debate

Multi-model LLM debate and second-opinion validation. Broadcasts a prompt to several models — HTTP providers or local CLIs — has them argue over N rounds, and returns a synthesized answer plus an agreement/disagreement report.