Last released Oct 11, 2026
Behavioural security layer for AI agents — watches every tool call your agents make, learns what normal looks like, and flags dangerous deviations