Last released Apr 19, 2026
Guardrails for AI agents — budget limits, kill switches, and human approval gates.
Supported by