Skip to content
Home

Kubernetes

One surface for every kube signal.

Cardinal pins Kubernetes events onto your metrics and logs, unfolds a pod's full topology on a click, and gives your agents a governed, read-only view of the cluster.

01 · Event overlay

See what the cluster did, the moment it happened.

Kube events (BackOff, CrashLoop, OOMKilled, Scheduled) pinned onto every metric and log chart. Hover a marker to see the pod, the message, and the exact moment against the signal.

  • · Auto-correlated by namespace, workload, and pod
  • · Works across metrics, logs, and trace charts
  • · No dashboard config; overlay is default on

Click to enlarge · esc to close

Kubernetes events pinned onto a metric chart in Cardinal, with a BackOff event tooltip open.

02 · Infra map

Click any k8s tag. Get the whole neighborhood.

Every Kubernetes entity in a metric legend, log field, or event is a doorway: pod, service, deployment, node, configmap, PVC. Click one and the topology unfolds around it, each entity coloured by current health.

  • · Works from any k8s tag Cardinal sees, not just pods
  • · Traversal is live, not a nightly-rebuilt CMDB
  • · Ten kubectl calls collapse into one graph

Click to enlarge · esc to close

Live Kubernetes topology in Cardinal: a service expanded into its deployment, pods, nodes, configmaps, and PVCs, each coloured by health.

03 · Server-side MCP

Give agents the cluster, read-only, on your terms.

A Kubernetes MCP hosted by Cardinal, with no local kubeconfigs and no per-agent creds. Agents in Claude, Cursor, and Codex investigate incidents against your live cluster through a central, governed endpoint.

Governance

  • Read-only capability surface: no apply, patch, or exec
  • Central RBAC per tenant, cluster, and namespace
  • Every call audited against the agent session that made it
  • One managed endpoint: no kubeconfigs shipped to laptops
Capability surfaceRead-only
  • list_api_resourcesDiscover kinds available in the cluster
  • list_resourcesEnumerate pods, deployments, services, nodes
  • get_resourceFetch a single object's spec and status
  • list_eventsKube events, filtered by namespace or reason
  • get_logsContainer logs, tailed or scoped by time
mcp.cardinalhq.io/kube

Get to the pod, the event, and the cause, in one place.

Point Cardinal at the telemetry your cluster already emits. The event overlay, infra map, and MCP are on the moment ingest starts.

Get started freeRead the docsDeploy walkthrough →