Radar is a Kubernetes UI that runs as a single binary with a built-in MCP server for AI agents.
Radar addresses the challenge of understanding cluster state and troubleshooting issues in Kubernetes environments. It works by running locally on your machine or in-cluster, connecting directly to the Kubernetes API without requiring agents, CRDs, or cluster-side installation. The tool provides real-time visibility through multiple views including topology visualization, resource inspection, Helm release management, GitOps integration, live service traffic analysis, cluster audit logs, and upgrade impact assessment. Updates stream to the browser via server-sent events, enabling responsive interaction even on large clusters.
Radar suits teams that want a self-hosted, privacy-first alternative to cloud-based dashboards. It works across any conformant Kubernetes distribution including GKE, EKS, AKS, minikube, kind, and k3s. The tool is particularly valuable for operators who need to understand cluster relationships and troubleshoot issues quickly, and for teams integrating AI agents into their operations through the built-in MCP server. It requires no account, no external dependencies, and keeps all cluster data local to your machine, making it suitable for airgapped or locked-down environments.
The project maintains active development with regular releases and responsive issue handling. The codebase is written in Go and designed for performance on large clusters with tens of thousands of pods. The tool includes comprehensive keyboard shortcuts and security features, with documentation covering all major views and use cases. The project welcomes contributions and provides clear development guidance for those wanting to extend or modify the tool.
