OpenCost

What it is

OpenCost is a cost monitoring and allocation tool for Kubernetes. It tracks real-time and historical spend by mapping cloud billing data (compute, storage, network, load balancers) to Kubernetes constructs — namespaces, deployments, labels, pods. It answers the question "what does this workload actually cost" instead of leaving teams staring at a lump-sum cloud bill. It runs as an in-cluster exporter, exposes a Prometheus-compatible metrics endpoint, and ships a basic UI for browsing allocation data.

Who builds it and why

OpenCost originated at Kubecost, which open-sourced the cost-allocation engine and donated it to CNCF, where it's now an incubating project. Kubecost still maintains the majority of the codebase and uses OpenCost as the free, open foundation under its commercial product (Kubecost adds long-term storage, multi-cluster aggregation, governance features, and support on top). This is a common CNCF pattern — vendor-led open core — and it's worth knowing going in: the roadmap is influenced by what Kubecost needs for its paid tiers, though the core allocation logic itself is fully open and functional standalone.

Production readiness signal

197 contributors and 6,615 GitHub stars indicate solid community engagement for a niche observability tool. Active commit history through mid-2026 and a release cadence into the v1.120.x line show it's not stagnant — this is a maturing, regularly patched project, not an abandoned one-off. CNCF incubating status means it's passed a baseline bar for governance, adoption breadth, and process maturity (higher than sandbox, below graduated). Apache-2.0 licensing removes legal friction for commercial use. Go is the right language choice for a Kubernetes-native metrics exporter — consistent with the rest of the ecosystem's tooling and easy to operate alongside kube-state-metrics or Prometheus.

Who should use this

• Platform teams that need per-namespace, per-team, or per-label cost breakdowns and don't want to hand-build them from raw cloud billing exports.
• Organizations running multi-tenant clusters where chargeback or showback to internal teams is a real requirement.
• Teams already running Prometheus/Grafana who want cost data as just another metric stream, queryable and alertable like anything else.
• Anyone evaluating Kubecost's paid product — running OpenCost first is the natural way to see if the core allocation model fits before paying for the extras.

Who should NOT use this

• Teams with a single small cluster and a handful of workloads — the operational overhead of running and maintaining another in-cluster component isn't worth it when you can eyeball the cloud bill directly.
• Organizations needing long-term (multi-year) cost history, budget forecasting, or anomaly detection out of the box — OpenCost's default retention and analytics are thin; you'd end up bolting on Kubecost or building your own storage layer anyway.
• Multi-cluster fleets needing a single aggregated cost view — OpenCost is cluster-scoped by default; cross-cluster rollups require additional plumbing or the commercial product.
• Teams wanting hard cost controls (budgets, quotas, enforcement) rather than visibility — OpenCost tells you what things cost, it doesn't stop spend.
• Non-Kubernetes environments — this tool is K8s-specific by design; if half your spend is VMs or serverless outside a cluster, you need a broader FinOps platform.

Alternatives

• Kubecost (commercial) — same lineage, adds multi-cluster views, long-term storage, and governance on top of the OpenCost engine.
• CloudZero — broader cloud cost intelligence platform covering Kubernetes and non-Kubernetes spend, less Kubernetes-native but stronger on org-wide FinOps reporting.
• Cast AI — combines cost visibility with automated rightsizing and autoscaling actions, useful if you want optimization, not just observability.

Pricing

Fully open source, Apache-2.0, free to run yourself with no feature gating. No paid tier of OpenCost itself exists — the commercial upsell path is Kubecost, a separate product.