Skip to content
NewKubeOn AI: a managed LLM gateway with cost per 1M tokens for every model

Platform overview

One agent per cluster. One hub for every dollar.

KubeOn joins what runs in your clusters with what you are billed for, then turns the result into allocations, budgets and changes your teams can act on.

Architecture

Data moves one way: out of your clusters

The agent reads the Kubernetes API and writes a snapshot. The hub reads snapshots and billing exports. Nothing calls into your clusters.

Your environment, outbound only
Clusters with the KubeOn agentEKS, AKS, GKE, OpenShift, on-prem
Cloud billing exportsCUR 2.0, Azure exports, BigQuery
FOCUS files and rate cardsAny provider or data center
Metricsmetrics-server and Prometheus
Object storage you ownS3, GCS or Azure Blob, KMS-encrypted

KubeOn hub

KubeOn Cloud or self-hosted in your account

  1. 1Ingest snapshots
  2. 2Normalize labels
  3. 3Allocate
  4. 4Reconcile
  5. 5Recommend
Allocation and chargebackBy team, namespace or label
Change proposalsPatches your pipeline applies
Budgets and alertsSlack, Teams, email, webhooks
Reports and APIScheduled, CSV, REST
Ask KubeOnAnswers with traced figures

Product

Eight areas, one cost model

Every area reads the same allocations, owners and policies, so numbers agree from one screen to the next.

Visibility: What it costs and who owns it

Action: What to change and who approves it

Platform: Answers and data you can trust

Data model

Owners from your labels, normalized at read time

KubeOn reads team, cost center and application from the labels you already use, then resolves spelling differences when it reads them. Nothing is re-tagged in your clusters, and mappings are effective-dated so history stays as reported.

How ownership works
Label keys and the key KubeOn resolves them to
Label in the clusterResolved key
Teamteam
teamteam
TeamNameteam
kubernetes.io/teamteam
cost-centercost_center
CostCentercost_center

Freshness

What happens, and when

Every hour, on the hour

The agent writes a snapshot of nodes, pods, workloads, volumes and services.

20 minutes past

The hub ingests new snapshots and checks each one against its row counts.

Hourly at :50

Budgets, alert rules and anomalies are evaluated and notifications sent.

Every 30 minutes

Savings are recomputed from the latest usage and prices.

Daily

Billing is read, the last 3 days re-read for late items, and costs allocated and reconciled.

Daily, after allocation

Change proposals are regenerated, keeping every reviewed decision.

Coverage

The same product on every platform

Billing sources and identities differ by provider. What you can do with the data does not.

KubeOn capabilities by platform
CapabilityAWSAzureGoogle CloudOn-premises
Cost allocation on max(request, usage)IncludedIncludedIncludedIncluded
Reconciliation to the billIncludedIncludedIncludedRate card
Savings and change proposalsIncludedIncludedIncludedIncluded
Node sizing and family changesKarpenter, node groupsNode poolsNode poolsNode pools
Budgets, alerts and anomaliesIncludedIncludedIncludedIncluded
Network transfer costIncludedIncludedIncludedRate card
Agent identityIRSA or Pod IdentityWorkload IdentityWorkload IdentityToken or service account
Self-hosted hubTerraform or HelmHelmHelmHelm, air-gapped

See your own clusters in KubeOn.

A 30-minute walkthrough on your billing data, with an engineer who has run Kubernetes cost programs.