Central hub
One place to understand what the system is doing right now — and what it will refuse.
Live activity
Watch verdict volume and trend shifts across the last 24 hours.
Enforcement health
Confirm guard nodes are up and latency is within expected bounds.
KPI reference
The top row of tiles summarizes system behavior over the last 24 hours. Use these metrics as a quick check before you drill into the feed and queue.
| Metric | Definition | Why it matters |
|---|---|---|
| Total Verdicts | Total decisions made in the last 24 hours. | A proxy for overall traffic through the control boundary. |
| Allow Rate | % of actions approved (ALLOW). | Sudden drops can indicate tighter policy, degraded context quality, or a surge in risky actions. |
| Held | Actions awaiting operator review (HOLD). | Indicates how often human-in-the-loop gates are being exercised; watch for growing backlog. |
| Blocked | Actions denied immediately (BLOCK). | Often maps to red lines; repeated blocks can signal compromised prompts, bad integrations, or policy misuse. |
| Avg Latency | Average time to produce a verdict. | Measures performance impact on agent workflows and operator responsiveness. |
Visual sections
Activity Monitoring
The 24h Verdicts chart shows verdict volume over time. Use it to spot spikes (sudden bursts of actions), shifts in verdict mix (more HOLD/BLOCK), and “quiet periods” that could indicate upstream outages or stalled automation.
Agent Health
Agent Health is where you verify enforcement reliability per guard node. Monitor uptime and responsiveness for each node, and investigate any node that shows sustained latency increases or missed heartbeats before it becomes an operator incident.