Dashboard
A high-level governance overview of agents, resources, policies, and activity
The dashboard is the landing page once agents are connected and making tool calls. Reasoning about agent governance comes down to five questions, and every block on this page answers one of them: who owns the agent, what it's doing, whether a given call was allowed, whether you can prove it after the fact, and whether you can stop it.
Who owns it
The Inventory card counts agents, resources, and policies (active / total) — the org's current footprint — plus the users who were active in the selected time range, which is a measure of activity rather than membership. For the owner and identity behind a specific agent or connected instance, see the Agent Registry.
What is it doing
Gateway traffic totals every request through the gateway; the tool executions trend chart and the recent activity feed break that volume down over time and event by event. The LLM usage card, cost chart, and top-models widget add token and spend trends once agents report usage through OpenTelemetry — until then these figures read zero and a card invites you to connect telemetry. When anomaly detection is enabled, an Anomalies card surfaces open findings and how many are high or critical, deep-linking into the triage queue for spend spikes, unusual tool volume, and other statistically unusual behavior.
Was it allowed
Gateway health is the share of tool calls that executed successfully, with a count of real errors such as timeouts and upstream failures — policy blocks are excluded, since those are intended outcomes, not failures. The blocked-call count badge on the tool executions chart and the top blocked tools widget show what your policies are actually stopping.
Can I prove it
The dashboard is a summary, not the record. Every decision behind these numbers — the actor, the policy rule that fired, and the result — lives in the audit log. For trends over a longer window than the dashboard shows, see Analytics.
Can I stop it
None of these are dashboard actions — they live on the pages the dashboard summarizes. Revoke an agent instance's access, or delete it outright, from its detail page (see Agent Registry); disable or reorder a policy rule to change what's allowed going forward, or delete a resource to remove it from every agent's tool list.