Kubecost · Stackwatch (acquired by IBM via Apptio) · infraplz.dev
Stackwatch (acquired by IBM via Apptio)
Kubecost
Kubernetes cost allocation platform that breaks cloud spend down to namespace, pod, and label level with real-time monitoring. ML-driven rightsizing recommendations analyze historical usage patterns to suggest optimal resource requests and limits, while anomaly detection catches cost spikes before the bill arrives. Available as open-core self-hosted and fully managed SaaS.
Kubecost is a Kubernetes cost allocation platform that attributes cloud spend to namespace, pod, deployment, label, and team granularity in real time. It reads kube-state-metrics, cAdvisor, and cloud billing APIs to build a cost model for every running workload, then applies ML rightsizing to 30-day P95 utilization patterns to recommend CPU and memory request reductions.
The open-core distribution provides full single-cluster cost visibility, rightsizing recommendations, and basic alerting via self-hosted deployment at no cost. Multi-cluster aggregation, long-term data retention, SSO, and advanced governance require the Enterprise tier.
Kubecost is the upstream project behind OpenCost, the CNCF specification for Kubernetes cost allocation. Acquired by IBM via Apptio. Supported clouds: AWS, GCP, and Azure.
Key Features
Real-time cost allocation: attributes Kubernetes spend to namespace, pod, deployment, service, label, and team at sub-minute granularity using kube-state-metrics, cAdvisor, and cloud billing APIs
ML rightsizing recommendations: analyzes 30-day P95 CPU and memory utilization per container and recommends resource request reductions with projected monthly savings
Network cost tracking: identifies cross-zone, cross-region, and internet egress costs attributed to specific pods and namespaces — a cost driver absent from most Kubernetes cost tools
OpenCost compatibility: upstream project behind the CNCF OpenCost specification, ensuring data portability and standards compliance with the broader ecosystem
Cost alerts and anomaly detection: configurable alerts for namespace spend thresholds and real-time anomaly detection for unexpected cost spikes before the billing cycle closes
Open-core self-hosted distribution: full single-cluster cost visibility and rightsizing are free; multi-cluster aggregation, SSO, and long-term retention require Enterprise
Integrations
7 total
orchestration
Kubernetes
incident
PagerDuty
messaging
Slack
ci / cd
Terraform
cloud
AWSGCPAzure
Pricing
2 tiers
Free
Free
Free tier: Up to 250 CPU cores or 50 nodes (SaaS), 15-day metric retention, community support
Enterprise
Contact sales
Unlimited clusters, unlimited metric retention, RBAC, SAML/SSO, multi-cloud, dedicated support
