Traces LLM reasoning to extract assumptions, calculations, conclusions, and flaws.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.