Run and debug

You can try a workflow at two sizes: one step at a time while you build it, and the whole workflow once the steps work. Both run against the real APIs through your integrations, so a step that sends an email really sends it.

Test one step

Select the step and click Test-run this step in its toolbar, or open Edit step and go to the Test tab.
Under Incoming state, choose what the step receives: click an earlier step’s chip to fill the override with its last saved test result, or expand Override the incoming state and type JSON.
Click Test (or ▶ Run). The Test also sends, automatically, the saved output and metadata of every earlier step whose saved test captured them, whether or not you click a chip. A {{ }} reference to metadata or output the saved test lacks reads as empty text; a { "$": ... } binding to it is refused and asks you to re-run that upstream test. A malformed expression is reported as a request error before the action runs.

How you know it worked: Test run: <step id> shows Result, the step’s Metadata, and any Console output. If it failed, it shows the error code and message instead.

The test uses the configuration saved on the step; only the incoming state changes between tests. To change what the step does, edit it on Configure first.

A step that calls a real app needs a passing saved test before the workflow can be published. Until it has one, the step shows “Not yet tested — a passing test is required before this step can be published.” Testing the step from here records that test.

For a trigger, the Test tab takes sample values for its fields and saves them as the workflow’s test input. Flow-control steps such as If and For each can’t be tested on their own.

Run the whole workflow

Make sure the trigger has sample values (see Triggers). Without them the run starts with empty fields.
Save your changes. Run runs the version saved on the server, not unsaved edits.
Click Run in the header.

How you know it worked: the button reads Queued…, then Running…, and each step’s border changes colour as it goes: accent while running, green when it succeeds, red when it fails. A dashed border means the step hasn’t run.

Run works on drafts, so you don’t need to publish first. Stop this run cancels a run in flight; steps already running finish first.

Read the run log

The run log opens beside the canvas while a run is in progress.

Errors are listed first: the run’s own error, then each failed step’s code and message.
Below that, one row per step, with a status (Pending, Running, Succeeded, Failed, Skipped, Queued or Canceled) and its timing.
Expand a row to see the step’s Input and Output, with every expression already resolved, and Metadata (app, version, action, connection) when the run recorded it. A failed step may show the metadata of its last attempt; a row with none recorded shows no block. Built-in (@w6w/*), Function and Workflow steps show no Metadata. Any input value that reads a vault secret, including a string with a secret inside it like Bearer {{ secrets.API_KEY }}, is shown as [redacted]. The step still sent the real value, and the run record and logs never store it.

The same per-step steps.<id>.meta also appears in the run results returned by ?wait=true, on the latest or a specific version’s run route (a wait that times out returns no steps).

Click × to close it. To reopen an older run, click Execution history in the header, pick the run and click Open in visual editor. The canvas then shows that run’s colours and log.

Troubleshooting

Symptom	Cause	What to do
`Fill the required fields to test.`	A required trigger field or parameter is empty.	Fill it in.
`The incoming state must be a JSON object to run.`	The override text isn’t a JSON object.	Wrap it in `{ … }`.
`This step is missing required configuration — set it on the Configure tab.`	A required parameter has no value.	Fill it on Configure.
`Run failed without reporting a reason.`	The run stopped before any step reported an error.	Open Execution history for details, and check the trigger’s sample values.
`steps.<id>.meta.*` reads empty in a step test	The earlier step’s saved test predates metadata.	Test that earlier step again.
A run uses old settings	You ran before saving.	Save, then click Run again.
Execution history is greyed out	The workflow hasn’t run yet.	Click Run once.

Where to next

History: every run in the project, with filters.
The tool rail: retry and failure handling for the whole workflow.
Publish and versions: publish once the steps pass.