Skip to main content
A plug is code that calls a system you do not control. Test it in layers, from your own code to a real run.

​
1. Test your own logic

Write the work of an action as plain functions, such as listOrders(token, query), and keep the action thin. Test the functions with any test runner against recorded answers of the other system. You then do not need Collabase to know that a date is parsed or a page is requested correctly.

​
2. Build and check

A successful build means the plug bundles into one file and that every dependency has an allowed licence.

​
3. Upload to a test instance

Upload to an instance that is not your production one, with an API token of an administrator there. Give each test upload a new version. See Publish and versions.

​
4. Test a step with real data

1

Create a flow

In a test workspace, create a flow and add a step of your plug.
2

Create a connection

If the plug needs credentials, create a connection with a test account of the other system.
3

Test the step

Fill in the inputs and select Test step. The step runs once in the sandbox, for real. The result is shown and saved, and the data picker of later steps offers its fields.

​
5. Read the run

Run the flow and open it under Runs. Check the input of the step as the plug received it, the output, and the message of a failed step. A secret never appears in the log: Collabase replaces the values of the connection with ***. See Debugging runs.

​
What to try

Webhook triggers cannot be tested before the flow is published, because the other system sends the data. See Webhook trigger.
Last modified on October 26, 2026