Skip to main content
Governance controls which tools agents may call and when a person must approve. LLM prompt/output safety lives under Guardrails (Studio LLM Governance). Use the workspace page for fleet analytics and the approvals inbox; use Graph Studio to attach policies to a specific Agent Graph.
Requires Professional or Enterprise (Pro+). Workspace OPERATE → Governance is Admin / Superadmin only. Studio read uses workspace.governance.read; create/attach need write permissions.

Setup guide

Create policy → attach → approval delivery → triage.

Tool policies

Allow, Human approval, Deny — wizard steps.

HITL & HIL New

Human-in-the-loop policies and destination profiles.

Approvals

Accept or Reject pending tool calls.

Guardrails

LLM safety (separate chapter).

​
Where in the product

​
Workspace tabs

Toolbar on Policies / Guardrails: Last 7 days / Last 30 days / Last 90 days · Refresh. Optional ?view=sessions drills into session history for policy or guardrail purpose.
Governance Policies dashboard

Workspace Governance — Policies tab

​
Policies analytics labels

Quick links: Human approvals · Observability · Billing & usage.

​
Studio Governance modules

Open an Agent Graph → Governance. Chrome title Governance. Tabs:
Governance Summary cards

Studio Governance Summary

There is no separate Studio Guardrails rail in this IA. Use LLM Governance inside Governance, or the docs chapter Guardrails.

​
Permissions and APIs (summary)