Test and iterate on your workflow in the new builder

After the new builder proposes a workflow plan, you test it by running each step with real data from your connected accounts. This article explains how to set up trigger data, approve or reject actions as the test runs, review step output, and handle issues so you can publish with confidence.

Beta

This feature is in beta. Some features may change as development continues.

Before you begin

• Complete the planning phase in the new builder. If you have not built a workflow yet, start with Create Zaps by describing your workflow.
• Make sure your apps are connected. The builder will prompt you to connect accounts during testing if any are missing.

1. Set up trigger data

Before the test runs, the builder fetches sample data from your trigger so it has real values to work with.

1. When prompted, review the trigger sample the builder has fetched. This is the data that will flow into your actions during the test.
2. If multiple samples are available, select a different one from the list.
3. If the sample does not look right, click Request new sample to fetch a fresh one.
4. If your trigger uses a webhook and has not fired yet, copy the webhook URL shown and send a test request from your source app.
5. If no sample data is available, you can either provide trigger data manually by entering values directly, or let the builder auto-generate test data. Review the generated values and confirm before proceeding.

Note

If the builder cannot fetch a sample, it will display a clear error category (configuration issue, missing authentication, no events yet, or a transient failure) so you know what to fix.

2. Approve actions during the test

As the test runs, the builder pauses at each write action and asks for your approval before executing it.

• Read actions (fetching data, searching records) run automatically without prompting.
• Write actions (creating records, sending messages, submitting data) require your approval each time.

For each write action, you can:

• Click Approve to run the action as proposed.
• Click Edit to adjust step parameters before approving. You can re-test the step after editing.
• Click Make changes to send instructions back to the agent and have it revise the step.
• Click Reject and provide feedback to have the agent retry with your input.
• Click Reject without feedback to have the agent try a different approach entirely.

You can also adjust how approvals work for the session in the execution settings:

• Turn on Auto-approve to skip approval prompts for the rest of the session.
• Turn on Pause after step to review every step's output before the test moves on.

Approvals inside paths

If your workflow has branching paths, the builder handles each branch independently. When a path's filter conditions do not match your sample data, you can choose to skip that path or test it anyway. Approving steps inside a path does not affect the rest of the test.

3. Answer clarifying questions

The builder may pause to ask for your input when it cannot determine a value on its own. Questions appear in these formats:

• Single-select: pick one option from a list.
• Multi-select: pick one or more options.
• Account picker: select a connected account or connect a new one inline.
• Field picker: a searchable, paginated list of available values for a specific field. Type to filter.
• Free text: enter a value directly.

You can skip any question if you do not want to answer it. The builder will note the limitation and continue.

4. Review step output

After each step runs, its output appears in an expandable card in the test timeline.

• Click a card to expand it and review the data that step produced.
• If Pause after step is on, the test waits for you to review before continuing.
• To change a step's parameters and run it again, click Edit on the card and update the values.
• To edit a step's output directly and continue from that point, update the values in the card and click Continue from here. Steps downstream of that point are cleared and the test resumes from the edited output.

If you disconnect during a test, the builder resumes automatically when you reconnect. Your progress is not lost.

5. Iterate on the workflow

If a step fails or produces unexpected results, you do not need to restart the test from the beginning.

• Provide instructions in the planning chat to ask the agent to change the trigger, swap an app, add or remove a step, or adjust how fields are mapped.
• The agent will apply the change and continue the test from where it left off. Steps that already passed are not re-run.
• If the same operation fails twice, the builder stops and explains the issue rather than retrying in a loop.

6. Check for publishability issues

As the test progresses, the builder flags issues that would prevent publishing, such as missing required fields or steps that exceed your plan's limits, directly in the timeline. Resolve these before moving to publish.

If your plan does not support a specific app or your workflow exceeds the step limit, the builder will block that step and show an upgrade link.

Next steps

Once your test passes, you are ready to publish.

• Create Zaps by describing your workflow
• The new builder