Self-managed install & maintenance > Install and config > Advanced config > Workload API > Enable the Workload API

Enable the Workload API¶

Deploy the Workload API, turn on its feature flags, and verify the result.

Deployed Kubernetes components¶

The following Kubernetes deployments are required for the Workload API to function:

Component	Description
`workload-api-app`	Workload API service—provides the control plane for artifact and workload CRUD, scheduling, and quota enforcement.
`workload-api-task-workers`	Async task workers—run the Workload API's background work: container builds, workload replacement, and the network-path readiness gate.
`covalent-cloud-server`	Compute scheduler—translates workload specifications into LRS resources.
`workload-monitor-app`	Event-driven agent—reports workload status back to the Workload API via Kubernetes pod watchers.
`tyk-api-gateway-app`	Tyk API Gateway—handles authentication, routing, rate limiting, and TLS termination.
`lrs-operator-manager`	LRS operator—provisions containerized workloads as Kubernetes Deployments.
`datarobot-otel-collector-app`	OTEL collector—receives traces and logs from running workloads.

Note

In its standard flow, the Workload API runs user-supplied, pre-built container images referenced by image URI—it does not build them, so the Image Build Service (IBS) is not required for this flow. (Building a workload from source is a separate Private Preview capability.)

Prerequisites¶

LRS operator installed and configured in your cluster.
datarobot-otel-collector chart enabled (see Deploy and enable).

Deploy and enable¶

Setting global.workload-api.enabled: true does two things: it deploys the workload-api, covalent-cloud-server, and workload-monitor components, and it sets the installation-level enablement (the ENABLE_* env vars) for the required feature flags. Enable the OTEL collector separately, as it is not included in the global flag:

global:
  workload-api:
    enabled: true

datarobot-otel-collector:
  enabled: true

The global flag sets the installation-level enablement for the following feature flags:

Feature flag	Description
`WORKLOAD_API_CONTAINERS`	Enables the Workload API feature.
`MLOPS_RESOURCE_REQUEST_BUNDLES`	Enables resource bundle sizing for workloads.

WORKLOAD_API_CONTAINERS gates a licensed feature—see Feature flags and licensing.

Feature flags and licensing¶

Setting global.workload-api.enabled: true enables the Workload API feature flags across the installation—there is no separate per-organization step to grant access.

The Workload API is included with the Agentic platform entitlement—it is not sold as a separate SKU and is not available on Predictive-only plans. Access is gated by two things:

Organization entitlement — the WORKLOAD_API_CONTAINERS feature flag ("Enable Access to Workload API for Image Containers"), enabled by default for Agentic and GenAI entitlement sets. It follows the standard GenAI feature-flag model; see the Generative AI feature flags reference for the flag list and prerequisite order.
Per-user seat — each user must hold an Agentic Builder Seat, allocated to the organization and assigned to users through the standard seat-license flow.

Seat provisioning uses the standard seat-license key (a DataRobot engineer generates it with the appropriate seat limits); contact your DataRobot representative or administrator. A System Administrator can optionally override the organization feature flag in the feature management console.

Verify the deployment¶

The 05_workload_api post-install check exercises the Workload API end to end—it creates a workload, waits for it to reach running, and cleans up afterward. The check skips gracefully when the agentic-seat license is absent, so interpret its result accordingly: a skip means the license isn't present (not a failure), a pass confirms the end-to-end path, and a fail indicates a real deployment problem.

To validate the deployment manually from a user's perspective, follow the end user documentation to deploy your first Workload and reach its inference endpoint.