kubectl MCP Server

by Rohit Ghumare

kubectl-mcp-server turns your AI coding assistant into a Kubernetes operator. 253 tools, 8 workflow prompts, 8 data resources. You describe what you want in natural language, and the MCP server translates it into kubectl commands, applies them, and reports back. Built by Rohit Ghumare, this is the most comprehensive Kubernetes MCP server available. It covers core operations (list pods, services, deployments, nodes), debugging (troubleshoot crashed pods, inspect logs), cost optimization (identify idle resources, right-size workloads), security auditing (RBAC analysis, network policy review), and Helm chart management. All through conversation. The real value is workflow prompts. Instead of stringing together 15 kubectl commands to diagnose a CrashLoopBackOff, you describe the problem: "My backend pod keeps crashing in the staging namespace." The server runs the diagnostic sequence, checks events, pulls logs, inspects resource limits, and surfaces the root cause. For experienced K8s operators, that saves 10 minutes per incident. For junior engineers, it bridges a knowledge gap that previously required a senior DevOps engineer on call. Installation is two lines: npx -y kubectl-mcp-server or pip install kubectl-mcp-server. Supports stdio (Claude Desktop, Cursor), SSE (web clients), and HTTP transport. Docker images available on both Docker Hub and GitHub Container Registry. Works with 15+ AI clients including Claude, Cursor, Windsurf, VS Code Copilot, and JetBrains AI. 859 GitHub stars and actively maintained. The project ships new tools regularly, with recent additions covering cost analysis and network diagnostics. If you run Kubernetes clusters and use AI coding tools, this is the MCP server to install first. Browse more MCP servers on Skila or read our guide to MCP and Model Context Protocol.

FAQ