KubeStellar is an Apache-2.0, CNCF Sandbox multi-cluster configuration management engine, written in Go, that lets teams deploy and configure workloads across many Kubernetes clusters — edge, multi-cloud, and hybrid cloud — using the same single-cluster tooling they already run.

What it is

KubeStellar is a platform for managing Kubernetes objects across multiple clusters. Its core orchestration engine handles workload propagation through BindingPolicy, WDS (Workload Description Space), ITS (Inventory and Transport Space), and WEC (Workload Execution Cluster) concepts. A developer defines a binding policy between clusters and Kubernetes objects, and KubeStellar uses regular single-cluster tooling to deploy and configure each cluster according to those policies. The project lives in the Kubernetes and Cloud Native Computing Foundation ecosystem, and it integrates with tools teams already use, eliminating the need to modify existing resources. It is accompanied by sub-projects: the console (an AI-powered web dashboard with 160+ cards, GPU monitoring, and AI missions), console-marketplace (153+ community card presets covering GPU/AI/ML, ArgoCD, OPA, Falco, and LLM-d), console-kb (an AI knowledge base of community missions and operational runbooks), and kubestellar-mcp (an MCP server providing AI agent tooling for Claude, Cursor, Windsurf, and VS Code).

The concrete problem it solves is per-cluster manual work. Without KubeStellar, multiple clusters are usually deployed and configured individually, which is time-consuming and complex. It replaces that individual deployment and configuration of each cluster with a single binding policy that drives propagation, giving a developer experience akin to using one cluster. The multi-cluster setups it targets exist for separation of environments such as development, testing, and staging; isolation of groups, teams, or departments; enterprise security and data governance compliance; resiliency across clouds; improved resource availability; heterogeneous resources; and running applications on the edge, including disconnected environments.

Key capabilities

• BindingPolicy definition that binds clusters to Kubernetes objects and drives propagation to each cluster.
• Workload propagation across the WDS, ITS, and WEC planes of the core engine.
• GitOps-oriented placement, rollout, scheduling, status, and summarization behaviour, as reflected in the project topic list.
• console dashboard with 160+ cards, GPU monitoring, and AI missions.
• console-marketplace with 153+ community card presets for GPU/AI/ML, ArgoCD, OPA, Falco, and LLM-d.
• console-kb knowledge base of community missions and operational runbooks.
• kubestellar-mcp MCP server exposing natural-language cluster operations to AI agents in Claude, Cursor, Windsurf, and VS Code.
• Support for edge and disconnected environments alongside multi-cloud and hybrid cloud clusters.

Who uses it and how

• Teams already deploying to a single Kubernetes cluster that want to expand to multiple clusters without rewriting resources.
• Teams already running multiple clusters that want a more streamlined developer experience than configuring each cluster separately.
• Organisations separating development, testing, and staging environments across distinct clusters.
• Enterprises that must isolate groups, teams, or departments, or satisfy security and data governance requirements.
• Operators running applications at the edge, including in disconnected environments, and those needing resiliency across clouds or access to heterogeneous resources.
• Platform teams that want cluster operations driven from an AI agent or a web dashboard instead of direct per-cluster access.

Getting started

The README points to a Quickstart in the documentation at docs.kubestellar.io. The Console can be tried at console.kubestellar.io in demo mode with no install required, and kubestellar-mcp ships its own installation instructions.

How it compares

This registry names no paid products that KubeStellar replaces, so no cost-model or licensing comparison is available here. Among tools named in the facts, ArgoCD, OPA, and Falco appear only as preset integrations inside console-marketplace rather than as comparable peers, and KubeStellar stands alone in this registry as a multi-cluster configuration management engine.

When to use it — and when not to

A self-hoster must operate a multi-cluster control plane covering the WDS, ITS, and WEC planes, with the Kubernetes clusters themselves as the backing infrastructure; the supplied facts name no required database, object storage, or SMTP dependency. Teams running a single cluster, or without Kubernetes operational experience, should not pick it, because the value only appears once several clusters must be kept consistent. Note the honest signals: 215 open issues against 711 stars and 316 forks, CNCF Sandbox rather than graduated maturity, and the repository is positioned as the core engine within a broader stack of sub-projects whose feature claims are documented separately.