The Autonomous, Recurring, Reviewable (ARR) framework characterizes an AI agent by three key attributes: its capacity for autonomous decision-making and action, its continuous operation through recurring processes, and its ability to review or be reviewed for performance and progress. This framework helps distinguish sophisticated AI agents from simpler, one-off AI interactions by emphasizing their operational independence and iterative nature.