Assign Roles to the Agent

Go to Administration > RBAC > Role Bindings and create a binding for the agent:

Field	What it means	Example
Subject	The agent's identity	ml-training-agent
Role	The role you want to assign	data-processor
Scope	An optional boundary to limit where the role applies	project:ml-pipeline

The role defines what the agent is allowed to do. For example, a data-processor role might allow:

• data:read — Read training datasets
• model:write — Write model artifacts
• logs:write — Write execution logs

warning

Only give the agent the permissions it actually needs. Avoid assigning broad roles like admin to autonomous workloads.

---

Next Step

Once roles are assigned, configure the allowed scopes on the client to control which permissions the agent can request.