Inference Hooks

An inference hook mechanism that Anthropic provides for Claude, allowing custom logic to be inserted at key points in model inference. By passing prompt and tool-call information through frames, it enables real-time safety interception and policy control.