MCP Server Kubernetes is an open-source Model Context Protocol server that lets AI assistants connect to a Kubernetes cluster and manage it through conversation. With about 1,600 stars on GitHub and active development in 2026, it is distributed as an npm package and loads your existing kubeconfig automatically.
Once connected, an MCP client such as Claude can inspect cluster resources, work with pods and deployments, and carry out management operations through the server’s tool interface, instead of you translating every request into kubectl commands by hand. The server supports loading kubeconfig from multiple sources in priority order, which makes it easy to point the same setup at different clusters in different environments. It is built for the emerging workflow in which an AI assistant acts as a copilot for platform operations: answering questions about cluster state, helping debug unhealthy workloads, and handling routine management tasks under your supervision, with the model explaining what it is about to do in plain language.
MCP Server Kubernetes is free and open source under the MIT license. There is no paid plan; your only costs are the cluster itself and the AI client you connect to it. As with any tool that can change cluster state, it should be run with credentials scoped to what you are comfortable delegating.
Platform engineers and developers who manage Kubernetes daily and want an AI copilot for inspection and routine operations, and teams experimenting with conversational infrastructure management. It suits users who already work with MCP clients and kubectl-style workflows.