Skip to content
Start free
Cost MonitoringCluster to container

Kubernetes cost for every cluster, namespace, workload and pod.

CPU, memory, GPU and storage priced per pod, with idle capacity and system overhead shared out. Drill from the whole organization down to one container.

Start freeBook a demo
  • 5-minute install
  • One Helm chart
  • First 100 vCPU free

Seven levels · one mental model

  • Organization

    All clusters

    $48,210

    mo · run rate

  • Cluster

    prod-eu-1

    $18,400

    mo · 38% of org

  • Node Group

    ng-general-large

    $7,032

    mo · 12 nodes

  • Node

    ip-10-0-3-42

    $586

    mo · m5.4xlarge

  • Namespace

    data-platform

    $4,820

    mo · 89 / 142 pods

  • Workload

    analytics-worker

    $2,140

    mo · 12 replicas

  • Pod

    78f9-3a1c

    $180

    mo · loaded total

Seven views

Seven views, from organization to pod.

Each has its own page with cost trend, usage and breakdowns. Pick a level to see it.

Now viewing · Organization

See the whole bill, all at once.

Run rate, MTD, savings opportunities, and resource efficiency across every cluster you operate. Rank clusters by spend, drill into the largest, find what changed since yesterday.

  • Org-wide KPIs

    Run rate, MTD, identified savings, efficiency.

  • Cost by cluster

    Ranked list with share, region, and pod count.

  • Spend by team

    Labels and namespaces rolled up org-wide.

The Organization view with sample data.

Cost Explorer

Group, filter, save and export.

Group by any of ten dimensions, filter by label or team, switch cost modes, save the view, and export CSV or PDF.

  • 10 group-by axes

    Workload, Namespace, Controller Kind, Node, Pod, Container, Deployment, StatefulSet, DaemonSet, Instance Type.

  • Composable filters

    Workload, node, and cluster axes, each with its own filter set. Pick label key=value, cluster, namespace, team, AZ, region, capacity type. Filters auto-prune when you change axis.

  • Two cost modes

    Fully Loaded includes each workload’s share of idle capacity and overhead. Workload Only shows direct usage. Switch between them in place.

  • Save, share, export

    Pin views as dashboards (private or shared with team). Export to CSV for finance, PDF for review. Daily, weekly, or monthly granularity.

Cost Explorer with sample data: 14 days of cost, grouped by workload.

Loaded cost

Idle capacity and overhead, charged to the pods that cause them.

Each pod’s cost includes what it uses, its share of idle node capacity, and its share of system pods. That is the number to compare with revenue.

  • Workload bill

    CPU, memory and GPU for the pod itself.

  • Idle attribution

    The pod’s share of unused capacity on the node it runs on, pro rata.

  • Overhead share

    System pods and cluster services, spread across the workloads that rely on them.

  • Loaded total

    The number to compare against revenue, capacity, and unit economics.

Pod · analytics-worker-78f9-3a1c

How $49.60/mo is built up

Loaded total

$49.60

Cost buildup · stacked

$49.60 / mo

  • CPU

    4.2 vCPU · us-east-1

    $14.00

    28.2%

  • RAM

    24 GB

    $7.00

    14.1%

  • GPU

    0.5 × A10G

    $24.00

    48.4%

  • Idle

    Allocated share

    $3.40

    6.9%

  • Overhead

    System pods

    $1.20

    2.4%

Why it matters

A pod that “costs $45” on paper actually costs $49.60 once you account for the idle node capacity it forces the autoscaler to keep, plus its share of cluster overhead. Kubeadapt computes that for every pod.

Cost dimensions

Cost over time, by resource, hour of day and pricing model.

  • Over time

    Run rate, month to date, trends.

    Daily, weekly and monthly views, with hourly resolution for ranges under 14 days.

    $66/hr

    current run rate

    Hourly
    7 days agotoday

  • By resource

    CPU, memory, GPU and storage.

    Each cost component broken out per cluster, namespace and workload, not one compute total.

    Component split

    4 components

    $48,210 / mo

    • CPU38%
    • Memory24%
    • GPU22%
    • Storage10%

  • Behavioral

    Hour-of-day patterns.

    See exactly when each workload peaks. Right-size your HPA bounds. Schedule batch into the cheap hours.

    Hour-of-day pattern

    14:00–17:00

    peak window

    Last 7d avg
    0006121823

  • Spot-aware

    Spot savings, surfaced.

    Spot coverage at every level: cluster, node group, namespace, workload. Find the gaps you can still close.

    Spot coverage

    72%

    of compute hours

    Saved

    $6,140

    / mo

    Spot72%
    On-demand28%

Install in 5 minutes

Stop paying for unused capacity.

Free up to 100 vCPU. After that, $1.99 per vCPU per month, dropping to $0.60 at volume.

Start freeBook a demo
Agent runs in your clusterNo credit card