Red Hat Advanced Cluster Management 2.17 opens fleet search to MCP clients

Red Hat Advanced Cluster Management for Kubernetes 2.17 adds a Technology Preview Model Context Protocol server that exposes the product’s fleet-wide resource search to AI assistants and other MCP clients. The release also adds OpenShift Lightspeed context attachment for managed clusters and Technology Preview integrations for Argo CD agent mode and Perses dashboards, according to the 2.17 release notes.

Fleet data becomes an MCP tool

The new server queries the same Red Hat Advanced Cluster Management search database that stores information about Kubernetes resources across managed clusters. Its find_resources tool supports filters for resource kind, name, namespace, cluster, status, labels, cluster selectors, text and age. Clients can request lists, counts, summaries or health views, then group, sort and limit the output, according to Red Hat’s MCP server guide.

Red Hat documents connections for both OpenShift Lightspeed and Claude Code. The Helm chart is installed on the RHACM hub cluster, discovers the MultiClusterHub installation and deploys the server with access to the search database. The documentation requires cluster-administrator access and labels the entire capability Technology Preview, so this is an evaluation path rather than a production support commitment.

For OpenShift Lightspeed, administrators add the server to the cluster’s OLSConfig resource. Claude Code connects through the route created by the MCP server installation and authenticates with an OpenShift token. Those examples make the design concrete: an assistant can ask structured questions about resources across a managed fleet instead of receiving a static documentation-only context.

Lightspeed also gets cluster context

Separately, RHACM 2.17’s console can attach managed-cluster details to an OpenShift Lightspeed conversation. From a managed cluster’s details page, the feature adds relevant YAML from the ManagedCluster and ManagedClusterInfo resources as context, the release notes say.

The distinction matters. Attach supplies context from the cluster page the operator is viewing; the MCP server provides a searchable interface across resources and managed clusters. Together they move Lightspeed from generic product guidance toward questions grounded in fleet state, while keeping both access and configuration under cluster administration.

More previews around GitOps and observability

The release also previews an argocd-agent mode that uses destination-based mapping for ApplicationSet deployment, reducing prerequisite configuration. In observability, RHACM can render hub and managed-cluster metrics in Perses dashboards through the Cluster Observability Operator console plug-in when the multicluster observability add-on UI option is enabled. Governance policies can now detect available minor operator updates on different channels, and virtualization administrators can distribute custom roles to managed clusters and assign them through MulticlusterRoleAssignment resources, according to the same release notes.

The practical headline is narrower than “AI manages the fleet.” RHACM 2.17 supplies a supported product release with an explicitly preview-grade bridge between fleet search and MCP clients. Platform teams can test whether that bridge improves investigation and inventory workflows, but should treat the MCP, Argo CD agent and Perses additions according to their Technology Preview status.