The Agent Node enables Journey Orchestration graphs to invoke AI agents directly within a workflow. Agents are goal-driven components that can interpret context, make decisions, use configured tools, and return structured outputs for downstream processing. The Agent Node allows orchestrations to combine AI-powered decision-making with deterministic workflow execution.
Benefits
Simplifies integration of AI agents into orchestration workflows.
Supports reusable, purpose-specific agents across multiple journeys.
Enables context-aware decision-making while maintaining deterministic orchestration control.
Supports separate Production and Non-Production environments for testing and deployment.
Prerequisites
Xponent Intelligence is enabled for your organization.
Contact Support team to enable Xponent Intelligence for your organization.
Agent Builder is available and enabled within Xponent Intelligence.
At least one agent must be created and available for selection before adding an Agent Node to a graph.
If Graph tools are used with agents, the project must include one Production environment and at least one Non-Production environment.
Creating an Agent Node
Here you can create an Agent Node to a graph to invoke an existing agent and use its output within your journey orchestration workflow. Before creating an Agent Node, create and configure the required agent in Agent Builder. Once the agent is available, you can create the Agent Node.
To create an Agent Node:
From the graph canvas, select the page, which opens the Create a New Item window.
Enter a name for the node.
From the Agent Node category, select Agent Node.
Click Create Item to create a new agent node, or click Cancel to cancel the process.
The Agent Node configuration page opens, where you can configure the agent, inputs, and outputs for the node.
.gif?sv=2026-02-06&spr=https&st=2026-09-10T10%3A46%3A00Z&se=2026-09-10T11%3A02%3A00Z&sr=c&sp=r&sig=7pxVk%2F9d%2FaeHz7hPZnAbV3vu9nwj9rZb2E1MmkjyO94%3D)
Agent
The Agent Node supports separate agent selection for Production and Non-Production environments.
Non-Production: Select the agent that should run in Non-Production environments.
Production: Select the agent that should run in the Production environment.
At runtime, the graph uses the agent configured for the target environment. The same agent can be selected for both Production and Non-Production environments if required.
Agent functionality supports environment-aware configuration. Projects using graph tools for agents must contain one Production environment and at least one Non-Production environment.
If a project has multiple environments, the Production environment uses the production agent for execution, while all other environments use the non-production agent for execution.
Non-Applicable Agents
The Non-Applicable Agents toggle is available in both the Production and Non-Production agent lists.
When enabled, the list includes agents associated with the opposite environment and not assigned to the currently selected environment. These agents remain visible for reference and display their assigned environment tag.
In the Non-Production list, enabling the toggle displays agents tagged as Production.
Those agents can not be selected and display one note: “This agent is in Production and cannot be selected as a Non-Production agent”.
In the Production list, enabling the toggle displays agents tagged as Non-Production.
Those agents can not be selected and display one note: “This agent is in Non-Production and cannot be selected as a Production agent”.
Search and Preview
Use the search box to quickly locate an agent by name.
To preview details of a selected agent, select the View (
👁️ ) icon next to the agent. A preview panel opens displaying the agent details.Select the Close (X) icon to close the preview.
Input
The Input section defines the runtime data that is sent to the agent. Input data provides the context required for the agent to analyze information, make decisions, and generate a response.
Input Schema: Select the required schema and configure the input fields that will be passed to the agent during execution.
Input Context: Use Input Context to provide additional instructions or supporting information that helps the agent interpret the input data correctly.
Output
The Output section defines the response generated by the agent and made available to downstream nodes in the graph.
Output Schema: Define the JSON structure that the agent must return. A well-defined output schema helps ensure the response can be processed consistently by downstream nodes and workflows. Agent Nodes are designed to return structured outputs that can be consumed by orchestration graphs.
Output Instructions: Output Instructions define how the agent should generate and format its response. Provide clear instructions to help the agent return consistent, structured output that aligns with the configured output schema and can be used reliably by downstream orchestration steps.
Output Destination: Specify where the generated output should be stored for use by subsequent nodes in the graph.
Saving Changes
After completing the configuration:
Select Save Edits to save the Agent Node configuration.
Select Discard Changes to remove any unsaved changes.
The status indicator displays No Unsaved Changes when all updates have been saved successfully.
After the Agent Node is configured and saved, the graph can invoke the selected agent during execution and use the returned output in subsequent workflow paths.
Adding an Agent Node to a Graph
After creating and configuring an Agent Node, you can add it to an orchestration graph and connect it to other nodes within the workflow.
To add an Agent Node to a graph:
Open the required orchestration graph. Right-click anywhere on the graph canvas and select Add Node.
The Add Node To Graph window is displayed. Scroll to the Agent Node section, which lists all available Agent Nodes.
If multiple Agent Nodes are available, use the search box in the Add Node To Graph window to quickly locate the required node by name.
Select the required Agent Node from the list and click Add Node.
The selected Agent Node is added to the graph canvas and is ready to be connected to the appropriate upstream and downstream nodes.
.gif?sv=2026-02-06&spr=https&st=2026-09-10T10%3A46%3A00Z&se=2026-09-10T11%3A02%3A00Z&sr=c&sp=r&sig=7pxVk%2F9d%2FaeHz7hPZnAbV3vu9nwj9rZb2E1MmkjyO94%3D)
After finishing the graph changes, deploy the graph to make the node available for execution. During runtime, the deployed configuration uses the selected environment-specific agent, processes the configured input data, and returns structured output for use by subsequent nodes in the orchestration workflow.
The Agent Node executes the agent configured for the target environment and returns structured output that can be consumed by downstream nodes in the orchestration workflow.
Best Practices
Test agent behavior, tool integrations, and output structures in a Non-Production environment to validate configurations prior to production use.
Agent executions can encounter issues such as timeouts or unexpected outputs. Use graph error handling to manage these scenarios gracefully.
Agent execution may take longer than traditional orchestration nodes. Design use cases with the expectation that responses can take several seconds to complete.
Agent Nodes work best when handling a specific decision or task instead of controlling an entire journey workflow. Smaller-scope agents are easier to understand, test, and troubleshoot.
Use Agent Nodes to enhance orchestration rather than replace it. Combine AI-driven decisioning with deterministic workflow logic for greater traceability and control.
Use Agent Nodes at clearly defined decision points and keep the graph in control of the overall workflow. Agent outputs should flow into deterministic paths for predictable orchestration behavior.
Well-defined instructions help improve response quality and reduce the likelihood of unexpected results.
Agent responses are not always repeatable; compliance-sensitive or auditable business logic should remain in deterministic orchestration paths.
FAQ
Can I test an Agent Node before deployment?
Yes. You can test an Agent Node during configuration to validate connectivity, inputs, outputs, and agent responses before deploying the graph.
Can I use an Agent Node in graphs that contain Batch or Segment Listeners?
No. Agent Nodes cannot be used in graphs that include Batch Listeners or Segment Listeners because high-volume scenarios may introduce latency and rate-limit constraints. Graph validation and deployment will fail if this combination is used.
Can I modify an existing Agent Node?
Yes. You can update Agent Node configurations, such as instructions, context, input mappings, and output definitions, and then retest the changes before deployment.
What happens if an agent does not respond in time?
Journey Orchestration enforces a 30-second node timeout for agent executions. If the agent does not return a response within the allowed time, the node may time out. Consider implementing deterministic fallback or error-handling paths in your graph.
How do I know if an agent succeeded or failed?
Validation and execution errors, such as missing fields or timeouts, are surfaced through graph execution. Agent execution logs are also available for troubleshooting and investigation.
Can Agents be used in low-latency journeys?
No. Agent execution may take between 5 and 30 seconds. Deterministic nodes are recommended for low-latency scenarios.
Can Agent Nodes be used for auditable or compliance-driven decisions?
No. Agent outputs are not always repeatable and are therefore not ideal for auditable or compliance-driven decision-making scenarios. For predictable and compliance-critical processes, use deterministic orchestration logic within the graph.