Know what every pod costs. Allocate by namespace, deployment or label. Rightsize requests/limits with utilisation-driven, HPA-aware recommendations — no in-cluster agent required.
By the numbers
EKS · GKE · AKS
0
control planes supported
CloudWatch CI · Cloud Monitoring · Azure Monitor
0
DaemonSets installed in your cluster
rightsize · idle deploy · oversized PVC · cluster · idle node
0
Kubernetes rules
SHOWBACK · CHARGEBACK
0
chargeback modes
How it works
EKS reuses the AWS read-only role + CloudWatch Container Insights. GKE uses the GCP service account + Cloud Monitoring. AKS uses the Azure service principal + Azure Monitor. Nothing is deployed inside your cluster.
Container Insights · Cloud Monitoring · Azure Monitor
KubernetesChargebackService allocates direct workload + PVC cost, then splits shared cluster/node cost by namespace request weight (`cpuRequestCores + memoryGiB/4`). Maps namespaces to teams/products via metadata or tags.
workload · PVC · shared split
Five rules surface waste: rightsize-workload, idle-deployment, oversized-PVC, over-provisioned-cluster, idle-node. Each shows the request vs the observed peak with a $/mo impact.
request vs observed peak
What's inside
Allocate node, network, storage and managed-service cost down to the pod. Roll up by namespace, deployment, label or — once mapped — by team. Export chargeback CSVs that finance can actually use.
Recommendations factor in autoscaler behaviour, restart frequency and SLO budget — so 'reduce CPU request to 200m' isn't going to wake you at 3 AM.
Identify which workloads can safely move to Spot or Preemptible nodes based on disruption tolerance, restart history and current SLO posture.
Capabilities
Before vs after Lumicost
Without Lumicost
With Lumicost
Where teams use it
Show every team a monthly invoice for their namespaces — including idle pod cost — to align incentives without micromanaging.
Find dev/staging deployments running with 0 traffic for 7+ days. Most teams cut 25–35% of their non-prod K8s bill on the first sweep.
Compare $-per-vCPU and $-per-GB-RAM across clusters and clouds — find the one whose node pool sizing actively works against you.
From the field
“We refuse to ship a Helm chart for cost visibility. Container Insights and Cloud Monitoring already have the metrics — your platform team has enough to maintain.”
Plays nice with your stack
FAQ
No. Lumicost reads cluster metadata via the cluster API and joins it with cloud-native usage metrics (CloudWatch / Cloud Monitoring / Azure Monitor). No exporter to maintain.
Cost allocation requires a billing source. For bare-metal clusters we accept a custom $/hr cost per node pool and allocate from there.
Connect read-only credentials. First insights in 24 hours. No credit card.
90 seconds · read-only credentials · no credit card