Guardrails text
An optional free-text policy prepended to every Agent’s system prompt. Use it to set ground rules that apply across all agents — for example, tone requirements, data handling rules, or escalation instructions.Saving changes
Click Save guardrails after editing. Changes take effect on the next Run — agents already running are not affected mid-run. Guardrails revisions are tracked in the revisions panel below the editor.Blocked domains
Blocked domains are now managed under Settings → Browsing. See Team Browsing for details.Related
- Team Browsing — Blocked domains
- Roles and Permissions — Who can view and edit guardrails
- Guardrails for High-Risk Automations — Best-practice patterns for HITL gates, caps, and allow/deny lists
- Browsing Capabilities — How agents browse the web