Real calls to an Agent5i deployment. The organization policy
governance_demo decides what runs.
checking…
Tools Agent5i itself runs (built-in tools, MCP servers) are checked by the platform before they execute. The app cannot skip the check.
Tools served through the SDK run in your app, but the platform checks each call
first. A blocked call never reaches the app; the model gets
policy_denied or approval_pending.
Functions your app runs on its own are checked only if the app asks
(client.governance.guard). The platform cannot stop an app that does not
ask. Every decision is recorded as evidence either way.
| Id | Tool | Run | Arguments | Group |
|---|