The Extension tool node

The Extension tool node in CloudRadial AutomationAI makes exactly one call to one tool of an installed extension, such as adding a note to a ticket, updating a ticket, or posting a message, with the arguments you set. No AI model decides whether or how to call it, so the node does the same thing every time it runs. This article is for workspace users with the Admin role or higher who build workflows in the Designer.

• What the Node Does
• Before You Start
• Configuring the Node
• Tools That Write to Other Systems
• Runner Version Requirements
• Output and Errors

What the Node Does

Add an Extension tool node from the Integrations section of the palette. When the step runs, AutomationAI resolves the node's arguments, checks that every required argument has a value, and sends the call to the deployment's runner. The runner calls the tool once and reports the tool's result back as the node's output.

The call runs on your runner, so the extension's credentials stay with your runner and never pass through AutomationAI. You write no code for this node; AutomationAI generates the step for you.

Use the Extension tool node when you know exactly which call to make. Use the Agent node when the step needs to reason about which tools to call. The Extension tool node works on the main path and in the On finish lane, where it is the usual way to report a run's outcome back to a ticket.

Before You Start

• The extension must be installed in your workspace, with its secrets set. See Browsing and installing extensions.
• The workflow must be deployed to a runner that can reach the extension's service
• A tool that writes to another system needs an Admin or Owner to approve it when the workflow is published or tested

Configuring the Node

Select the node to open its settings:

• Name: the node's display label on the canvas
• Extension: one of your installed extensions
• Tool: one of that extension's tools. The tool's description appears below the picker.
• Arguments: one field per argument the tool accepts, generated from the tool's own definition. Required arguments are marked with an asterisk. If the tool takes no arguments, the editor says so.

Each argument accepts either a literal value or a binding expression. Type a value directly, or use the Bind… menu beside the field to insert a binding from an earlier node or from the run itself. For example:

{{ nodes.trigger.output.TicketId }}
{{ run.summary }}

For the binding syntax, see Passing data between nodes: input and output JSON. For the run.* values, such as the run's status, failures, and summary, see the article Reporting a run's outcome with the On finish lane.

Changing the extension or tool clears the arguments, because each tool defines its own.

Tools That Write to Other Systems

A tool that changes data in another system, such as creating a ticket note, is a write tool. When you pick one, the node and its settings show a Writes badge. Once approved, a write tool runs without asking each time, so AutomationAI requires an approval before it can run:

• When an Admin or Owner publishes the workflow, or runs it with the Test action, the Designer opens an Approve calls that write to other systems dialog that lists every write call in the workflow. Select I approve these write calls. and confirm to publish or run the test.
• A Member cannot publish or test a workflow that contains a write call. The Designer reports that only an Admin or Owner can do it.
• The approval is recorded on the published version, along with who approved it and when. Publishing a new version asks again.
• If a tool that was read-only when the version was published later becomes a write tool, the step fails with a message that the tool was not approved when the version was published. Publish a new version to approve it.

Read-only tools, such as lookups, need no approval.

Runner Version Requirements

Read-only tool calls work on any current runner. Approved write calls require runner version 2.3.0 or later; the current runner release is 2.4.1. Runners that update themselves pick up new versions automatically. If a runner is older than 2.3.0, a write call fails with a message asking you to upgrade the runner, unless another instance of the same runner has already upgraded, in which case that instance picks up the step. To upgrade an older runner, see Upgrading a runner to a new version.

Output and Errors

• The tool's result is the node's output, available to later nodes at {{ nodes.<node-id>.output.result }}
• If a required argument resolves to nothing, the step fails before the tool is called, naming the missing argument
• If the tool returns an error, the step fails with the tool's error message. As with any failed step, its branch stops there and the run ends Failed, unless the step is in the On finish lane.
• Each call has a two-minute time limit

For help with a failed step, see Troubleshooting a failed or timed-out run.

If you are still having trouble, we're here to help! Submit a ticket here for assistance, and don't forget to check our status page to ensure there are no outages in your area.