Create and Configure Workflow Agent Teams for Use in Agentic Apps

To build agentic apps with agent teams, first configure Workflow type agent teams for optimal results. Only published workflows can be used in agentic apps.

Best Practices

• Use an agent team of type Workflow. Workflow agent teams enable structured inputs and predictable flow.
• Make sure that the description of your agent team clearly defines the agent team's capabilities and specialization. When multiple agent teams are engaged, this description is used to help identify which team is best suited to handle a given query.
• Design smaller, topic-focused agent teams. You can combine multiple agent teams as needed within the agentic app.
• Keep workflows lean to reduce latency. Limit the number of LLM and agent nodes where possible to reduce perceived delay when resources are requested from the agent team.
• Pass user input and history explicitly. In workflows, conversation history and user input aren’t automatically passed to internal LLM, agent, or workflow nodes. Use a code node or variable assignment to pass what you need (for example, using $context.$system.$inputMessage), and then reference those variables in subsequent LLM or agent calls.
• When possible, end your workflow with an agent or LLM node to help prevent output streaming.