Enables context-loop navigation in agents by exposing goto and loop_start_marker tools that delegate to a host over a Unix socket, allowing bookmarking and jumping between context iterations.
Enables AI models to maintain a session-scoped checklist of todos, reading and updating it via todo_read and todo_write, with optional JSON file persistence.
Enables agents to analyze their session context usage and surgically remove heavy or unnecessary events from their history, preserving coherent transcripts and freeing token budget without restarting.