Product
See which namespace and workload needs attention. Open its pod status, restarts, events, logs, requests, limits and routing without piecing them together from separate tools. Fortem runs locally using your existing kubeconfig.
Local install · read-only first · no Helm chart
Fortem normalizes the affected workload, recent change, Kubernetes events, safe log-pattern counts and available traffic metrics into a compact local assessment.
Fortem treats a namespace as an environment in the first release, then normalizes workload, pod, resource, and event data into one scan-friendly view.
Open an environment, then inspect Deployments and StatefulSets, their pods, container images, rollout details, restart counts, and waiting or termination reasons.
Investigate an OOMKilled container →Requests and limits come from Kubernetes. Actual CPU and memory usage appears only when a compatible metrics source is available; missing telemetry is never rendered as zero.
Related Kubernetes events and container logs live beside workload details, with filtering and clear source labels. Readiness is shown as a fact, not a promise that the application is healthy.
Inspect Ingress, Service, and backend relationships from Kubernetes. HTTP request, error, and latency metrics require a separate compatible metrics source and retain their measurement point and time window.
Understand what request counters actually measure →Fortem starts read-only. Supported restart and scale actions require an explicit startup flag, an exact object confirmation, and a Kubernetes authorization check before the API accepts them. An assessment never authorizes a mutation.
Nodes show scheduler-visible whole-GPU and MIG capacity beside Pod requests. Fortem Pro can project allocation cost from a rate you provide for each resource type; it does not claim measured GPU utilization or a cloud bill.
Data contracts
Optional integrations enrich the same interface. Their absence becomes “unavailable,” with a reason—not a misleading zero.
The default incident engine is deterministic, runs offline, and records evidence, caveats and provenance. It does not send cluster data to Fortem.
The same standard API path supports EKS, GKE, AKS, kind and k3s: contexts, namespaces, workloads, pods, nodes, services, ingresses, events, logs, requests and limits.
Current CPU and memory usage. The rest of the interface continues to work when it is unavailable.
Five-minute request count, RPS, 4xx, 5xx, and p95 latency from a compatible ingress-nginx, Traefik or Istio profile.
On clusters with a readable Karpenter API, NodePools show provisioned CPU and memory capacity against configured limits. Without it, Fortem keeps the base node view and labels the source unavailable.
Fortem has a limited AWS/GCP/Azure node-rate estimate, not provider billing. Pro can model GPU unit-rate cost when you supply the price. The GPU figure is separate and must not be added to the node estimate.
Install with your coding agent, start in synthetic demo mode, then select a kubeconfig context you control.
Local install · read-only first · no Helm chart