We are at re:Invent 2026Book my merch
Cloud Intelligence™Cloud Intelligence™

Kubernetes

Every Kubernetes dollar has an owner. Tags optional.

Tagging tracks less than half of cloud spend. Cloud Intelligence™ combines your cloud bill with in-cluster utilization and runtime traffic to put every node, namespace and workload on the team, customer or feature that used it.

One Helm install · EKS, GKE, AKS and self-managed · Included in every plan

prod-eks · us-east-1 · last 30 days● 0 SLO breaches
1WorkloadsPodfit1,284 pods rightsized · CPU requests −41%$31,400
2NodesInfrafit96 → 61 nodes · bin-packing 58% → 83%$22,900
3RatesCommitments74% of node hours on Savings Plans and Spot$14,800
Saved per month, all three layers
$69,100 · 47% lower
Workloads 45%Nodes 33%Rates 22%
Network · seen by the sensorCross-AZ traffic flagged$6,240/moTop drivercheckout-api → redis

Trusted by teams running Kubernetes at scale

Akamai
Blumira
Chess.com
Current
OneFootball
Promptly
Recurly
Lusha
Mux

you can't label your way to 100%

Namespaces and labels stop at the shared stuff. We don't.

Shared clusters, node pools and the databases behind them have no clean labeling surface. Cloud Intelligence™ reads runtime traffic and allocates every dollar by what actually happened. Your namespaces and labels still count, as one more signal.

See how runtime attribution works
What the bill contains · where tags reach
Compute (tagged)
Tag
Managed DBs
Partial
Shared Kubernetes
Runtime
Network, NAT, egress
Runtime
Kafka, queues, caches
Runtime
LLM tokens, GPUs
Runtime
Self-managed on EC2
Runtime
Reachable by tags
43%
Reachable by Cloud Intelligence™
100%

every cluster, every cloud

One Kubernetes view, from bill to fix

Cost & Usage reports

Drill from cluster to namespace, workload and container, with CPU, memory and GPU.

Pod waste and idle nodes

Overprovisioned pods and idle nodes priced, per cluster, from real utilization.

Insights

Unused clusters, underused node pools and version upgrades, flagged automatically.

Kubernetes Intelligence

EKS, GKE, AKS and self-managed clusters, side by side.

Rightsizing recommendations

Recommended cost and resiliency risks per workload, from PerfectScale.

Budgets, alerts and CloudFlow

Guardrails and automations on the same numbers.

live in minutes

Connect Kubernetes in three steps

1

Connect your cloud accounts.

Enable Kubernetes core and we discover clusters daily. Add self-managed ones by name.

2

Install the agent and sensor.

Helm charts for the utilization agent and the runtime sensor. Data within 24 hours.

3

Every dollar gets an owner.

Shared clusters and nodes allocated by real usage.

Read-only agent. Kubernetes Intelligence is included in every plan.

Start free

Frequently asked
questions

Which clusters are supported?

Amazon EKS, Google Kubernetes Engine, Azure Kubernetes Service and self-managed clusters. EKS and GKE costs come from your cloud bill, AKS uses billing data when it is available, and self-managed clusters use an estimate based on list prices.

What does the agent collect?

CPU, memory and GPU requests, limits and usage per container, workload, namespace and cluster. The agent is the PerfectScale exporter, installed with one Helm command.

How fresh is the data?

The agent starts transmitting within a few minutes of install, and data can take up to 24 hours to appear in Kubernetes Intelligence. Cluster discovery on connected AWS and Google Cloud accounts runs daily.

Can I allocate Kubernetes costs without labels?

Yes. The runtime sensor reads traffic to assign shared clusters, and the databases and network behind them, to the team, customer or feature that used them. Namespaces and labels still count as one more signal.

Does Kubernetes Intelligence cost extra?

No. Kubernetes Intelligence is included with every Cloud Intelligence™ plan. PerfectScale for Kubernetes automation is included with Enterprise, or available as its own subscription with a 30-day free trial.

Do you support GKE Autopilot?

Yes. The runtime sensor is certified for GKE Autopilot. GKE Intelligence supports Autopilot with some limits, because GKE cost allocation doesn't cover every Autopilot resource.

Connect Kubernetes. Find the owner of every dollar.

Free to start. Read-only. Live in minutes.