Command (Dashboard)

Command is the home surface of the KLA Control Plane, served at the route /dashboard. The KLA Control Plane is a govern-in-place runtime safety, audit, and governance layer for enterprise AI agents: you instrument your existing agents instead of re-platforming them. Command is where everything those agents do becomes visible in one place. It answers three questions at a glance: What needs my attention right now? How are my agents trending? What is happening this second? From here you move directly into the module that resolves whatever you found.

Who uses it

Command is the daily landing page for everyone with a stake in agent operations. Platform operators watch live runtime health and active incidents. Compliance, risk, and audit officers check governance posture and how much evidence is sealed and ready. Developers and integrators confirm that a freshly instrumented agent is reporting spans and passing policy. It is deliberately role-aware: the same surface foregrounds different work depending on who is looking.

Key capabilities

Command organizes information into three operational tabs plus a row of summary cards.

Triage. The Triage tab is the priority queue for things that are wrong or stuck. It surfaces active safety alerts, Escalations awaiting a human decision in the Decision Desk, Assurance Alerts flagging drift or quality regressions, and failing Rollouts (a Rollout is a deployment of an agent Release). Each item carries enough context to act, including the affected agent, the policy or check that fired, and a one-click jump to the module that resolves it. Triage is designed to reach empty: when it is clear, your governed estate has nothing demanding intervention.

Analytics. The Analytics tab shows cumulative trends rather than point-in-time status. It charts active agent counts, request and token volumes, spend in genai.cost.usd, policy trigger rates broken down by outcome (allow, warn, require_approval, block), and validation results over time. These trends are the early-warning system: a climbing require_approval rate or a cost spike on one agent shows up here long before it becomes an incident, and each chart links through to the underlying Lineage Records.

Live Runtime. The Live Runtime tab is a real-time stream of system activity: in-flight requests, agent and collector heartbeats, and the append state of the cryptographic evidence ledger. Operators keep it open during a high-stakes Rollout or an incident, confirming second by second that agents are responding and that evidence is still being recorded.

The summary cards sit above the tabs as an always-visible posture strip:

How it connects

Command is a router, not a dead end. Every signal it shows is owned by a downstream module, and every card and Triage row is a link.

flowchart LR
  CMD["Command /dashboard"] --> DD["Decision Desk"]
  CMD --> ER["Evidence Room"]
  CMD --> AC["Assurance Center"]
  CMD --> LE["Lineage Explorer"]
  DD -->|approve or reject| CMD
  AC -->|new Assurance Alert| CMD
  ER -->|bundle sealed| CMD

• Decision Desk owns the Pending Approvals card and the Escalations in Triage. When a policy returns require_approval, execution pauses and an Escalation routes here for a human to approve or reject.
• Evidence Room owns the Evidence Ready card. Spans flow through the KLA Collector and into a cryptographic ledger, where they are assembled into Sealed Evidence Bundles for download.
• Assurance Center owns System Posture and feeds Assurance Alerts into Triage when an agent drifts from its expected quality or fairness baseline.
• Lineage Explorer is the drill-down target from any chart or alert: click through to the exact Lineage Record (the full trace of an agent run) behind the number.

Start your day in Command, clear Triage, confirm posture, and use the cards as a map into the rest of the KLA Control Plane.