Finding an app version and deploying it to a specific workload cluster are normally two different tools and two different mental models. `organization_validate_access` is the quiet one worth knowing about — it answers the permission question before a create fails halfway.
A Go MCP server for the Giant Swarm App Platform. It runs over stdio against your current kubeconfig context and covers the app lifecycle: browsing catalogs, creating and updating apps, handling their configuration, and deploying to workload clusters through the CAPI integration.
- Apps — `app_list` with filters, `app_get`, `app_create`, `app_update`, `app_delete`
- Catalogs — `catalog_list`, `catalog_get`, `catalog_refresh`, `catalog_search`
- Catalog entries — `appcatalogentry_list`, `appcatalogentry_get`, `appcatalogentry_versions`, `appcatalogentry_search`
- Configuration via ConfigMaps and Secrets — `config_get`, `config_create`, `config_update`, `config_values`
- Organizations — `organization_list`, `organization_namespaces`, `organization_info`, `organization_validate_access`
- Workload clusters — `cluster_list`, `cluster_get`, `cluster_apps`
- System — `health`, `kubernetes_contexts`
- Resources addressable by URI: `app://{namespace}/{name}`, `catalog://{name}`, `config://{namespace}/{app}/values`
Go 1.21 or later to build from source with `make build`, then run the binary — it speaks stdio, so any MCP-compatible client can launch it. Access to a Giant Swarm management cluster and a valid kubeconfig; the server uses your current context unless you point `KUBECONFIG` elsewhere. Organization-based namespaces are supported directly.
Build from source — clone the repository and build it, then point your client at the binary