KubeLedger

KubeLedger is a self-hosted Kubernetes accounting tool that measures CPU, memory, and GPU consumption by namespace. It suits platform teams that need internal chargeback, showback, and capacity planning without a software bill. The trade-off is that your team operates the deployment and must govern cluster permissions, while its optional AI interface has limited first-version safeguards.

Overview

RealOps Labs maintains KubeLedger, a project originally created and architected by Rodrigue Chakode. It requires read-only access to Kubernetes API endpoints, including the Metrics API, and supports installation on Kubernetes and OpenShift through Kustomize or Helm. That keeps accounting inside your environment and avoids dependence on a hosted service, but cluster operations, upgrades, and permission management remain your responsibility.

The accounting model includes non-allocatable resources and node overhead instead of counting only workload consumption. This gives chargeback and showback figures a more realistic basis. Allocation is at namespace level, which works for teams assigning costs to applications or departments by namespace; organizations that need ownership split below that boundary will need another model.

Key features

Accounting and right-sizing

KubeLedger tracks CPU, memory, and GPU usage over time and compares consumption with resource requests. That combination helps expose over-provisioned workloads and supports capacity decisions rather than merely reporting historical usage. Rightsizing recommendations are included, although the accounting granularity remains namespace-based.

Retention and dashboards

The service retains a full year of usage analytics while maintaining a fixed footprint under 100 MB. Built-in views cover hourly CPU and memory trends for seven days, daily accounting for 14 days, and monthly accounting for 12 months. The different windows make short-term troubleshooting and longer planning possible, though the dashboard horizons are fixed rather than user-defined.

Exports and monitoring

Charts and datasets can be exported as PNG, CSV, or JSON. A Prometheus exporter and official Grafana dashboard provide familiar routes for teams that already centralize Kubernetes telemetry. These options make KubeLedger more useful in an existing observability stack than a dashboard-only accounting tool.

AI access

An optional MCP server exposes ten read-only tools to MCP-aware assistants such as Claude, Gemini CLI, Cursor, Windsurf, and Cline. It is disabled by default, but version one has no authentication at the tool layer and exposes the service as a ClusterIP. Teams should treat that as an integration for controlled internal environments, not as a ready-made secure assistant gateway.

Pricing

KubeLedger has one published plan: KubeLedger (self-hosted) costs free. Internal use is free and source code is available, but the software cannot be sold as a managed service. There are no paid seats or usage quotas stated for this plan, so it is a straightforward fit for teams willing to run it themselves. The restriction matters to consultancies and vendors that want to package KubeLedger as a hosted offering.

Platforms

KubeLedger is delivered through an API, a self-hosted deployment, and a web interface. Kubernetes and OpenShift installation via Kustomize or Helm suits infrastructure teams already managing those platforms; it is not positioned as a standalone desktop application or managed cloud service.

Who it's for

Choose KubeLedger when namespace-level accounting, internal chargeback or showback, and rightsizing are the priority, especially if a year of retention and Prometheus/Grafana integration matter. It is also a sensible option for organizations that require source-available, internal deployment at no software cost. Look elsewhere if you need a vendor-operated service, allocation below namespaces, or stronger built-in controls for AI-tool access.

Pros and cons

Pros

• Complete cost basis: non-allocatable resources and node overhead are included, producing more credible internal accounting.
• Useful planning history: one year of analytics, plus hourly, daily, and monthly dashboard windows, supports both investigation and capacity planning.
• Operationally light footprint: the fixed footprint is under 100 MB.
• Open integration paths: PNG, CSV, and JSON exports, a Prometheus exporter, and an official Grafana dashboard fit existing workflows.
• No internal-use software charge: the self-hosted plan is free and provides source code.

Cons

• Namespace-only allocation: teams requiring team, project, or workload splits below namespaces will find the model limiting.
• Self-hosting burden: installation, upgrades, API permissions, and ongoing operation sit with your organization.
• MCP security limits: the optional server lacks tool-layer authentication and is exposed as a ClusterIP in version one.
• Managed-service resale is prohibited: the free plan cannot be sold as a managed service.

Alternatives

Kubernetes Cost Management Software is the broader category to browse when you want to compare several accounting and optimization approaches.

Pick KubeAdapt instead if a freemium product with a free allowance covering the first 100 vCPU, all features, and email support is preferable. Atmosly Kubernetes Cost Optimization is a better fit for a small, bounded setup: its free Personal plan covers one cluster, one cloud account, one seat, one project, two environments, and ten applications. Choose PerfectScale when a free tier of up to 200 vCPU per month with unlimited clusters, nodes, and pods matches your scale.

Krossboard is another free, self-hosted API and web option. OpenCost is the clearer choice for an Apache 2.0 open-source project that requires Kubernetes and Prometheus. Podledger suits users wanting a forever-free community plan with single-cloud deployment and pod-level cost visibility. CloudBolt is aimed at broader cloud-management needs, with a free core CMP allowance for up to 100 resources. Klustercost is an alternative open-source project under GPL-3.0.

Verdict

KubeLedger is a strong choice for Kubernetes or OpenShift platform teams that want free, self-hosted namespace accounting with realistic overhead treatment, year-long retention, and established Prometheus/Grafana routes. Its main reasons to look elsewhere are the operational work of running it, the namespace-only allocation boundary, and the MCP server's limited access controls.