Calculates LLM prompt and completion costs across 14 models and suggests cheaper alternatives.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.