Compresses LLM context while preserving key information and cutting token usage.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.