Open Cluster Management

You have an MCP Server that acts as a gateway for GenAI systems to interact with multiple Kubernetes clusters. It enables multi-cluster resource operations, observability, and centralized control through the Model Context Protocol (MCP).

How to use

Use the MCP client to connect to the MCP Server and perform cross-cluster operations. You can retrieve resources from the hub cluster (current context), access resources from managed clusters, connect to a specific managed cluster, and interact with multi-cluster resources such as policies and add-ons through MCP. You can also fetch and analyze metrics, logs, and alerts from integrated clusters to gain end-to-end observability. When you start using the server, you operate through the centralized MCP interface to manage and observe multiple Kubernetes clusters from a single point.

How to install

Prerequisites you need before installing are: a machine with Node.js and npm installed, and kubectl configured for your clusters.

Create the MCP server configuration file with the provided snippet and use it to start the server.

{
  "mcpServers": {
    "multicluster-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "multicluster-mcp-server@latest"
      ]
    }
  }
}

Additional notes

Note: kubectl is required and should be installed on your system. By default, the MCP server uses the KUBECONFIG environment variable to access clusters. In a multi-cluster setup, it treats the configured cluster as the hub cluster and accesses others through it.

Configuration and security

The server configuration shown defines a single MCP server entry that runs via npx to fetch the latest multicluster-mcp-server package. Ensure you maintain proper access controls for the hub and managed clusters. If you use environment variables for authentication or cluster access, propagate those securely to your runtime environment.

Troubleshooting

If you encounter connection issues, verify kubectl can reach all clusters from the hub context, check that KUBECONFIG is pointing to the correct file, and confirm the MCP server container/process is running. Review logs for any authentication or RBAC related errors when attempting to access managed clusters.

Available tools

fetchHubResources

Retrieve resources from the hub cluster (current context) to provide a consolidated view of the environment.

fetchManagedResources

Retrieve resources from managed clusters to enable cross-cluster awareness and management.

connectToCluster

Establish a connection to a specific managed cluster using a given ClusterRole to authorize access.

multiClusterAccess

Access resources across multiple Kubernetes clusters through the Open Cluster Management gateway.

observeClusterMetrics

Retrieve and analyze metrics, logs, and alerts from integrated clusters to support observability.