Writing node scripts with the inline AI assistant

Script nodes in the CloudRadial AutomationAI Designer use a full code editor, and an inline AI assistant can draft the script for you from a plain-language description. You can then try the node on its own with test input before running the whole workflow. This article is for workspace users with the Admin role or higher.

• The Code Editor
• Generating a Script With AI
• Testing a Single Node

The Code Editor

Open a PowerShell Script, Python Script, or C# Script node in the Designer to edit its code in the built-in editor. The editor matches the node's language and is where you read input and emit output — see Passing data between nodes: input and output JSON for the input and output calls.

Generating a Script With AI

Each script node has an inline AI assistant beside the editor. To draft a script, describe what the step should do in the prompt box ("Describe what this script should do…") and select Generate. The assistant generates code in the node's language — PowerShell, Python, or C#.

To ground the result against the tools you have, add installed extensions under Context extensions before generating. The assistant returns its code as a diff against the current script:

• Select Accept & apply to write the generated code into the editor
• Select Discard to leave the editor unchanged

Testing a Single Node

You can run one node in isolation before testing the whole workflow. With a node open, select Test node. The test panel opens for that node.

To feed the node sample data, enter a JSON object in Trigger input (optional JSON). That input seeds {{ nodes.trigger.output.<field> }} for the test, so the node can read it the same way it would read a real trigger payload. If the box is empty, the node runs without trigger input; if the JSON is invalid, AutomationAI prompts you to fix it before running.

The test runs the current draft on a selected deployment's runner, so the workflow needs at least one active deployment. Pick a deployment and select Run test; the panel streams the node's result, which you can read to confirm the script behaves as intended.