Skip to main content

​
Two Levels of Control

ArgentOS gives you control over your agent’s behavior at two levels, depending on your needs: Both systems use the same underlying intent engine. Safety Rules is a simplified interface that writes to the same config. You can start with Safety Rules and upgrade to full governance when you need it.

​
Safety Rules (Core)

Available to all ArgentOS users. Set basic guardrails for your agent in under a minute. Read the Safety Rules guide Three controls:
  1. Cautiousness level — How careful should your agent be? (Relaxed, Balanced, Strict)
  2. Never do — Actions that are always blocked, regardless of context
  3. Requires approval — Actions where your agent will ask before proceeding

​
Intent Governance (Business)

For teams running multiple agents across departments with compliance requirements. Read the Governance guide Full hierarchical policy engine:
  • Global > Department > Agent inheritance (children can only tighten, never loosen)
  • Monotonic rules — neverDo is additive, allowedActions is subset-only, escalation thresholds get stricter
  • Simulation gates — Test agent behavior before deployment
  • Industry packs — Pre-built configurations for law, medical, financial, MSP, and more
  • Department templates — Support, Engineering, MSP Ops, Research

​
How They Relate

Safety Rules writes enabled, runtimeMode, globalPolicy.neverDo, and globalPolicy.requiresHumanApproval. The Governance Console exposes the full hierarchy. Both read and write the same intent config block.