Structures AI conversations into hierarchical thread trees, replacing full history with concise summaries to reduce token consumption and preserve context.