Features
Everything the cluster knows, on one screen.
Metrics, logs, events, topology and cost — read straight from the API server, correlated before you ask. No scrape configs, no agents, no query language to learn.
01 — Observability
Signals arrive already joined to the workload that produced them.
Every panel is one click from the pod, node, and release behind it — the context you'd otherwise assemble by hand across four tools.
Built-in metrics
CPU, memory and disk from the Metrics Server, per pod, node and namespace — with 24h history.
CPU13%
Memory15%
Disk4%
Logs & events
Stream any container, filter cluster events by type, reason or object, and keep both side by side.
22:58 helm upgrade payments
23:04 OOMKilled · payments
23:12 CrashLoopBackOff
Topology graph
Ingress to service to deployment to pod to node, drawn from live owner references.
ingress/web
svc/web · 3 endpoints
deploy/web · 3/3 ready
4 pods · 2 nodes
02 — Resource coverage
A page for every built-in resource.
Not a YAML viewer with a search box — real pages with the fields operators actually check, plus raw manifests one keystroke away.
WorkloadsPods · Deployments · StatefulSets · DaemonSets · ReplicaSets · Jobs · CronJobsRestarts, probes, images, rollout status
ConfigConfigMaps · Secrets · Resource quotas · Limit ranges · HPAsSecrets masked by default
NetworkServices · Ingresses · Endpoints · Network policiesTLS expiry and host routing
StoragePersistentVolumes · Claims · Storage classesBound / pending, capacity, mode
Access controlService accounts · Roles · Bindings · Cluster rolesWho can do what, resolved
ClusterNodes · Namespaces · Events · Helm releases · CRDsHealth score and capacity
03 — Against the usual stack
One binary instead of four moving parts.
K8Lens is not a replacement for long-term metrics storage. It is a replacement for the four tools you open before you find the answer.
The usual stack
—Prometheus + node exporters to install and tune
—Grafana dashboards to build and maintain
—An agent per node, and its contract
—kubectl, still, for the actual answer
K8Lens
+One container, ~60MB, starts in under a second
+Reads the API server with your kubeconfig
+Nothing installed inside the cluster
+Root cause written in plain language
See it against your own cluster.
One docker run. Read-only by design.