Flux MCP Server

GitOps failures are rarely visible in one place: the source may be stale, the Kustomization may be suspended, and the HelmRelease may be reporting an error from three reconciliations ago. This server gives a model the pieces to correlate them — Flux instance status, arbitrary Kubernetes resources, pod logs, cluster events and metrics — plus write tools to reconcile, suspend or resume a resource once the cause is understood. Because it can switch kubeconfig contexts within a session it can diff how the same application is configured in staging and production. A documentation search tool is included so the model grounds its answers in the current Flux API rather than an outdated apiVersion it remembers.