The deployment operator

The deployment operator (deployment agent) is a thin controller installed on each workload cluster in your fleet. It polls the Console for services to apply, runs stack and sentinel jobs locally, maintains the auth-proxy websocket when enabled, and reconciles agent-scoped CRDs on that cluster.

This is distinct from the management controller, which runs only on your management cluster and registers fleet-wide resources with the Console.

Agent configuration

Per-cluster tuning — poll intervals, concurrency limits, image registry overrides for job pods, and websocket behavior — is configured with the cluster-scoped AgentConfiguration CRD. See AgentConfiguration for the required singleton name and full spec reference.

Global settings that affect the entire fleet (including default agent helm values) are configured on the management cluster via DeploymentSettings.

API reference

Agent-scoped CRDs are documented in the Agent API Reference. Go types live in the deployment-operator API.

Full working example

The plural up command is the simplest way to generate a full repository from the ground up. It includes terraform for the management cluster, workload clusters, and bootstrap manifests for both the management controller and deployment agents. A public reference is available in our demo repo.