mcp-server-kubernetes

What it does

The server wraps kubectl to expose Kubernetes cluster operations as MCP tools. It loads cluster authentication from your kubeconfig file (defaulting to ~/.kube/config) and provides tools for querying, listing, describing, and creating Kubernetes resources. This lets Claude interact with your cluster directly—inspecting pod status, examining resource configurations, and deploying workloads.

Who it's for

DevOps engineers and platform teams managing Kubernetes clusters who want to automate operations through Claude. SREs troubleshooting cluster issues interactively. Anyone integrating cluster management into broader AI-assisted infrastructure workflows.

Common use cases

• Query pod status and logs to debug failing deployments
• Describe Kubernetes resources to understand configuration and diagnose issues
• Create or update deployments, services, and ConfigMaps from Claude
• Check cluster health and resource usage across multiple kubectl contexts
• Integrate Kubernetes operations into AI-driven infrastructure automation

Setup pitfalls

• Security: The scanner detected 8 secrets in the repository. Review the source code before deploying to production to ensure no hardcoded credentials are bundled.
• kubectl must be installed and in your system PATH. Verify with kubectl version.
• Kubeconfig file must exist and contain at least one configured cluster context. Test your setup locally with kubectl get pods before connecting the MCP server.
• Service account permissions: your cluster's RBAC policies must grant the account appropriate roles for the operations you intend (e.g., get, list, create on Pods, Deployments, etc.).