Last released Jul 23, 2026
A per-action autonomy policy layer for AI agents: decide what an agent may do on its own, based on an action's reversibility and blast radius rather than the model's capability.