AgentCouncil runs the same prompt through several AI coding agents on your project (currently using the pi harness), with any mix of providers and models, and saves each agent's final response as a Markdown report. A consolidating agent then combines the reports into a findings report of consensus findings, unique findings, and contradictions, ready to paste into an AI chat, and a companion report that assesses each agent's work and lists its duration, token usage, and cost.