Workflow Testing

Run, inspect, and debug a workflow before promoting it to production.

Workflow testing runs your workflow against a real or mock payload from inside the editor, so you can confirm each node behaves correctly before any production event triggers it. You feed in an input, execute the workflow (or a single node), and inspect the resolved input, output, and any errors at every step.

Use testing while you build to catch broken templating, bad mappings, expired credentials, and unhandled edge cases early.

Before you test

You need two things in place before a test run executes:

A Linked Account for the integration

Tests run against a real Linked Account so any API calls inside the workflow hit the actual third-party app. Pick one from the Testing panel under the Prerequisites section.

If the account is unauthenticated or its token has expired, the editor shows a warning. Click Authorize Application to reconnect.

Once the account is connected, the status changes to Connected.

A sample input payload

Every workflow needs an input to run. The editor auto-loads a sample payload based on the workflow’s trigger (event, webhook, schedule, or API call). Edit the JSON to match the shape of the payload you expect in production.

Use Reset to restore the auto-generated sample.

Click Save after configuring the Linked Account and payload. The workflow is now ready to test.

Run a test from the workflow editor

You can test a single node in isolation or execute the full workflow end-to-end.

Test a single node

Use this when you want to verify one node’s logic without running everything upstream.

1. Open the node

   Click the node you want to test in the canvas.

2. Switch to the Input/Output tab

   Inside the node’s side panel, open the Input/Output tab.

3. Click Run Node

   Press Run Node. Refold resolves any variables referenced by the node using the sample payload and previously-tested upstream nodes, then executes that node against the Linked Account.

   A checkmark appears on the node once it succeeds. The resolved input, the raw output, and any error returned by the call all render in the same panel.

Single-node testing is not supported for Group nodes or paginated loops. Test these by running the full workflow instead.

Test the entire workflow

1. Open the Testing panel

   Click the Testing button in the editor side menu.

2. Confirm the payload and Linked Account

   Verify the prerequisites section shows a connected account and the payload you want to send in.

3. Click Execute Workflow

   Press Execute Workflow. Refold runs the workflow start to finish and writes a Run log capturing the input and output of every node.

4. Inspect the run

   Open the Run log to step through each node’s resolved input, output, and timing. Switch between past runs from the same panel to compare executions.

Mock the trigger payload

You rarely have a clean production event handy while building. To simulate one:

• Edit the auto-loaded payload directly in the Testing panel. The shape mirrors what the trigger would send at runtime, so you can replace IDs, change field values, or add extra keys before running.
• Reset to the default to get a fresh template if your edits drift too far.
• Save multiple payloads if your workflow supports it, so you can re-run the same fixture later.

Override variables and environment values

If your workflow reads from project-level variables or secrets, those values still apply during a test run. To test against different values:

• Update the variable in your project settings before running, then revert.
• Hard-code the value into the test payload and reference it from there for the duration of the test.

A first-class “override for this test run only” UI may not exist yet. Confirm with your Refold contact before relying on one.

Inspect node-level results

Every test run produces a per-node breakdown. For each node you can see:

Field	What it shows
Input	The fully resolved payload sent into the node after all variables and templates evaluate.
Output	The raw response the node returned (API response body, function return value, mapped object).
Error	The error message and (where available) the upstream HTTP status, retry count, and stack trace.
Timing	How long the node took to execute.

Open results in two places:

• On the node itself, via the Input/Output tab, for the most recent run.
• In the Run log opened from the Testing panel, for any historical run during this editing session.

Re-run with a previous input

To replay an earlier test:

1. Open the Run log

   From the Testing panel, switch to the run you want to replay.

2. Copy the input

   Copy the resolved input from the trigger node back into the payload editor.

3. Execute again

   Click Execute Workflow to run with the same input.

If your workflow editor exposes a Re-run button on past runs, use that instead of copying the payload manually.

Promote a tested workflow to production

A passing test does not by itself activate the workflow for production events. To ship it:

1. Resolve any pending changes

   Make sure all node-level edits are saved. Unsaved changes do not publish.

2. Publish the workflow

   Click Publish (or the equivalent action) in the editor. Refold promotes the current draft so that incoming trigger events run against it.

3. Verify in production

   Trigger one real event and check the run in your workflow run history.

Publishing replaces the live version. If you need to roll back, keep a copy of the previous workflow JSON or use Refold’s version history if available.

Testing patterns and fixtures

A few patterns that hold up as workflows grow:

• One payload per code path. Keep a short library of input payloads, one for each branch your workflow takes (success, missing-field, rate-limited, deleted record). Run each before publishing.
• Test downstream first. When debugging a multi-node failure, run the failing node in isolation with a hand-crafted input. This isolates whether the bug is in the node or in upstream data.
• Use Log nodes as breakpoints. Drop Log nodes between steps to checkpoint intermediate values during a full-workflow run.
• Wrap risky calls in Try/Catch. Test the catch branch by feeding a payload guaranteed to fail (bad ID, missing field). See Try/Catch nodes.
• Mock external IDs explicitly. Don’t depend on auto-generated IDs from your Linked Account’s sandbox: hard-code stable IDs into the test payload so re-runs are reproducible.

Common gotchas

Node shows a checkmark but the workflow still fails end-to-end

A node-level checkmark only means that node ran successfully with the sample input you provided. The same node can still fail when upstream nodes produce a differently-shaped payload at runtime. Always run a full Execute Workflow before publishing.

Variables resolve to empty strings or `undefined`

Templated variables reference outputs of upstream nodes. If you haven’t run those upstream nodes yet (or you cleared their output), references resolve to nothing. Run each upstream node once, or run the full workflow, to populate the values.

Linked Account auth expired mid-test

Tokens stored on the Linked Account can expire between test runs. If a node returns a 401, reopen the Testing panel and reauthorize before re-running.

Group nodes and paginated loops don't show output

Single-node testing isn’t supported for Group nodes or pagination. Run the full workflow to exercise these.

Test runs hit real third-party APIs

Because tests use a real Linked Account, they create real records, send real emails, and consume real rate limits. Point your Linked Account at a sandbox or test workspace whenever the integration supports one.

The payload editor rejects valid JSON

Trailing commas and single quotes aren’t valid JSON. If the editor refuses to save, paste the payload through a JSON validator first.

Related

Workflow overview

The fundamentals of workflows, triggers, and nodes.

Templating and variables

Reference upstream node outputs in your inputs.

Linked Accounts

Authenticate the third-party account your tests run against.

Log node

Capture intermediate values during a run.