Agent policies
Configure policy for connected agents
Define model access, tool permissions and usage limits, then enforce them through the integration points used by the workflow.
Implementation review
Effective policy
Review organization defaults and any narrower overrides before enabling an agent.
Enforcement
Verify how the connected agent handles allow, block and warn decisions. Confirm failure behavior and access to the policy service.
Change review
Record who changes policy and validate the effect on representative workflows before wider rollout.
Product scope
Zentinelle AI is an independently adoptable product with an MIT-licensed core. Support and implementation are scoped separately. Workbench seats do not automatically include every agent integration or operating service.
Review the events and enforcement available in the chosen integration before relying on a control. Workload location and model destinations remain separate architecture decisions.
Review current security and compliance information