A Kubernetes MCP server that skips the kubectl wrapper

What it does

This Model Context Protocol (MCP) server lets AI tools manage Kubernetes and OpenShift clusters. Written in Go, it exposes cluster operations—resource CRUD, pod logs and exec, Helm charts, and Tekton pipeline runs—as tools that assistants like Claude or Copilot can invoke directly. It reads your existing kubeconfig and works across multiple clusters simultaneously.

The interesting bit

The README insists this is not another kubectl or helm wrapper. Instead of shelling out to CLI commands, it is a native Go implementation that calls the Kubernetes API server directly, which it claims eliminates external dependencies and reduces latency.

Key highlights

• Native Go binary for Linux, macOS, and Windows; also distributed via npm, PyPI, and container images.
• Operates on generic Kubernetes and OpenShift resources with full CRUD, plus specialized support for Pods, Namespaces, Events, Helm, and Tekton.
• Multi-cluster support using standard kubeconfig files, with automatic detection of configuration changes.
• Optional OpenTelemetry tracing and metrics, including a /stats endpoint for real-time statistics.
• Configurable via CLI arguments, TOML files, or environment variables, with an extensive test suite.

Verdict

Worth a look if you want AI agents to manage clusters without installing a toolchain of CLI wrappers on the host. Skip it if you only need ad-hoc kubectl access and don’t use MCP-compatible assistants.