⌘K
Root Cause

Agent Activity

TrustLayer observability over the agentic reasoning. Queries, tools, latency, cost, confidence calibration, human override rate.

Guardrails active
Investigations run (30d)
138
Tool calls (30d)
4,921
Avg latency / investigation
42s
Avg cost / investigation
$3.18
Human override rate
12%
Hypotheses dismissed by review
23%
Guardrail blocks (30d)
8

Confidence calibration

Predicted vs actual validation rate, by confidence bucket. Well-calibrated near the diagonal.

Live agent activity

  • 2m ago
    reasoning
    INV-2049 · Holdout test complete · precision 0.91
    $0.42
  • 14m ago
    tool
    MES.query on Chennai Line 2 shift log (INV-2049)
    $0.08
  • 38m ago
    hypothesis
    INV-2050 · Hypothesis promoted: compressor supplier variance
    $0.61
  • 1h ago
    guardrail
    Blocked auto-action: containment requires human approval
    $0.00
  • 2h ago
    reasoning
    INV-2053 opened from pattern PAT-2054
    $0.55