Policies are declarative rules that monitor and flag agent behavior. Get alerted when agents exceed limits or exhibit unexpected patterns.

Policy Types

Spend Cap

Alert when an agent exceeds a cost threshold

Retry Protection

Flag excessive retry loops

Duration Limit

Alert on long-running executions

Token Limit

Flag high token usage

Error Rate

Alert when error rate exceeds threshold

Tip

Start with a Spend Cap policy to prevent runaway costs from unexpected agent behavior.

Policies vs Governance Hub

Policies are automated ingest-time rules and violations. Human review queues, AI system registry, incidents, and compliance evidence exports live in the Governance Hub.

Creating Policies

Go to Policies in the app (Govern section)
Click Create Policy (or go to Policies → create)
Select the policy type
Configure thresholds and scope (e.g. by agent)
Save; violations will appear on the dashboard and in Policies