1. How do I use AI Agent Studio?
  2. Add Multi Agent Nodes in AI Agent Studio

Add Multi Agent Nodes in AI Agent Studio

You can use Multi Agent nodes in which a supervisor agent routes requests to specialized worker agents.

Configure each worker agent with a specific responsibility and use the supervisor prompt and output specification to control routing and response consistency. In the node's properties, you can define how the supervisor agent evaluates incoming requests and selects the appropriate worker agent.

  1. Go to AI Agent Studio and open the Workflows tab.
  2. Edit a workflow and add a Multi Agent node to it.
  3. Enter a name and description for the node. The code is automatically populated based on the name.
  4. Select an error handler node to define failure handling.
  5. Specify the maximum number of interactions the node can run.
  6. Specify whether the agent must respond in the user’s language. Otherwise, the workflow default language is applied.
  7. Specify whether the agent must use the hallucination guardrail prompt every time it runs.
  8. Specify whether to include chat history as context for the supervisor and worker agents.
  9. In Agent Persona and Role, enter the supervisor agent's role, tone, and behavior instructions.
  10. In Prompt, use expressions to specify the routing decisions. For more information, see Expressions in AI Agent Studio.
  11. Select and add topics to define the areas covered by the Multi Agent node.
  12. Specify whether to enable summarization to generate a conversation summary after routing completes. If enabled, use expressions to specify the summarization prompt.
  13. In LLM, select a language model.
  14. Define the output specification for downstream steps and expression tools.
  15. Add Agent nodes within the Multi Agent node as worker agents.
  16. Enter a name. The code is automatically populated based on the name.
  17. From the Agent list, select an existing agent as worker agent for the Multi Agent node. The properties of the required fields are populated with the values defined in the selected agent.
  18. Publish the workflow.

When the workflow runs, the supervisor agent receives the user request, evaluates the intent, and routes the task to the most appropriate worker agent. The selected worker agent performs the task and returns the response. If a request spans multiple topics, the supervisor agent can coordinate across multiple worker agents and return a consolidated response.