Features · Guardrails
Boundaries you define.
Topics to avoid, actions that need approval and phrases that trigger a human handoff.
- AI disclosure
- Always on. Every persona says it's an AI.
- Topic limits
- Stay on purpose, decline the rest.
- Escalation
- Hand off to a person on your triggers.
Guardrails · example
guardrails: [{ avoid: "medical advice", escalate: true }]More features
ManavOS · Guardrails