Red Hat is developing an open source Model Context Protocol (MCP) server that lets AI assistants like VS Code, Copilot, and Cursor interact with Kubernetes and OpenShift clusters via natural language. Written in Go as a single binary with no external dependencies, it talks directly to the Kubernetes API rather than wrapping kubectl or Helm. It supports RBAC restrictions, read-only or non-destructive modes, and can run locally or in-cluster over Streamable HTTP or SSE. It joins several existing Kubernetes MCP implementations from Stacklok, independent developers, and Microsoft's Azure-focused version.
Questions this post answers
What is Red Hat's Kubernetes MCP server and how is it different from other Kubernetes MCP implementations?
It is an open source Model Context Protocol server that lets AI assistants such as VS Code, Microsoft Copilot, and Cursor interact with Kubernetes and OpenShift clusters through natural language. Unlike many other implementations, it is not a wrapper around kubectl or Helm; it is a Go-based native implementation that talks directly to the Kubernetes API server, ships as a single binary with no external dependencies, and can run locally or in-cluster over Streamable HTTP or SSE. Teams weighing AI-driven cluster tooling can follow Kubernetes MCP server developments on daily.dev.
How can administrators restrict what an AI assistant is allowed to do through a Kubernetes MCP server?
Access can be scoped using standard Kubernetes RBAC, for example by creating a dedicated service account such as cluster-reader, and the server can be set to read-only or non-destructive mode so it only retrieves events, logs, and resource usage without executing changes. By default the server allows full CRUD operations on any resource, including custom resources, so locking it down is an explicit choice. Admins securing AI-assisted cluster access can track RBAC and MCP safety practices on daily.dev.
What other open source MCP servers exist for working with Kubernetes besides Red Hat's?
Stacklok offers a Go-based implementation called mkp, and MCP K8S Go is another Go-based option. The mcp-kubernetes-server project is Python-based and wraps kubectl actions and the Python Kubernetes SDK, developer Alexei Ledenev packaged an MCP server in Docker with kubectl, helm, istioctl, and argocd support, and Microsoft maintains an Azure-specific MCP Kubernetes implementation with built-in Azure security privileges. Developers comparing Kubernetes MCP tooling options can follow the ecosystem's growth on daily.dev.