A real Microsoft Agent Framework agent on Azure AI Foundry models, governed in-process by Cedar via GovernedAgent. Click a prompt below and watch policy decide what it's allowed to do.
Each prompt runs through the same governed path. Green should be allowed; red should be blocked or corrected by policy.
Three of the four governance categories, live: action authorization (always on), input guardrails and output evals (⚙ = enable the check/judge on this agent in the control plane). The fourth — the governance loop — is where these decisions get reviewed, captured to a dataset, and re-checked when policy changes, shown in the control plane.
lookup_order, get_forecast, execute_shell, lookup_incident, create_incident, update_incident, delete_incident, bulk_close_incidents, delete_incident_declared, get_forecast_declared