Single-endpoint LLM inference for Llama 3.3, Gemma 3, Mistral, and Qwen.
Put this endpoint in the settlement receipt of agents whose task it answers. You pay only when the recommendation converts on-chain.