kubectl-mcp-server

npx -y kubectl-mcp-server

{
  "mcpServers": {
    "kubectl-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "kubectl-mcp-server"
      ]
    }
  }
}

The interesting design choice here is not the raw tool count but the detection tools that sit in front of each add-on group. Before an agent starts asking about canary rollouts or backup schedules it can call the matching detect tool and find out whether Argo Rollouts or Velero is even installed, which is what keeps a question about a bare cluster from turning into a series of failed API calls. Coverage runs from ordinary pod, deployment and RBAC inspection through Helm chart and repository management to cost tooling that flags idle and overprovisioned workloads. It runs over stdio for desktop clients or SSE and streamable HTTP when it lives in the cluster, and the project is listed in the CNCF landscape.