Preview

Introduction

The Preview screen lets you test and validate your agent right after you build it. It gives you an interactive space to run the agent and review its output in detail. It has two tabs: Execution and Result.

Execution

The Execution tab is where you start a test run.

Configure inputs and run the agent

Set the values your agent will use, then trigger the run.

View the agent node viewer

The Agent Node Viewer shows what happens behind the scenes when your agent runs. It lays out each part of the workflow, including tasks, prompts, tools, enhancers, and the language model, and shows how they connect. This makes it easy to follow how data moves through the agent as it works through your inputs.

The node borders are colour-coded so you can read the status at a glance:

Inspect the request and response of each node

Click any node in the viewer to open its details and review both the request sent to it and the response it returned. This is especially useful for troubleshooting, since you can see exactly where a problem starts and confirm that each node behaved as expected.

Result

After a run finishes, switch to the Result tab to review the outcome.

Final content result

The processed response the agent produced.

Request payload

A detailed view of the data sent to the agent during the run.

Full responses

The complete set of responses returned during the run, for when you need the full detail behind the final result.