Testing Agent

Testing Agent uses your app in a real browser, the way a visitor would, and tells you which flows work and which don’t. When a flow fails, you get the reason in plain words and a fix with one click, all before you publish.

Testing Agent works alongside Security Scanner in the Studio. Security Scanner tells you if your app is safe, and Testing Agent tells you if your flows work the way you expect.

Where to Find Testing Agent

Open your project in the Studio. You can start Testing Agent from two places:

• The Testing Agent tab, next to Preview, Edit, Code, Cloud, Security and Settings. Look for the circle-check icon.
• The Publish window. Until your project has tests, you see Check before you publish there. On paid plans, click Set tests to start setup.

When your project has no tests, the Testing Agent tab opens on Publish knowing it works. Click Set up tests to get started.

Setting Up Tests

Click Set up tests. You don’t need to write anything to get started.

1. Testing Agent switches you to Preview and scans your project for key flows.
2. A card in the chat suggests up to six flows. To get more, ask the agent for a number, up to 20.
3. Tick the flows that matter to you.
4. Click Run tests to save and run them now, or Add tests to save them and run them later.

What a test includes

Each test has three parts:

• A short title, like “Contact form submits and shows a thank-you message”
• The test plan: the steps the agent follows, written so someone who has never seen your app could follow them
• The expected outcome: the result you can see on screen that means the test passed

In the Testing Agent tab, each test shows its title with the test plan under it. You can change any part of a test at any time.

Edit or delete a test

1. In the Testing Agent tab, open the test’s More actions menu and choose Edit test.
2. The agent shows the test in the chat and asks what you want to change.
3. Tell it what to change. You can change the title, the test plan or the expected outcome.

“Change the expected outcome to: the confirmation page shows the booking date.”

An edited test shows Not tested yet until you run it again.

To remove a test, choose Delete test from the same menu. Its past runs stay in Scan history.

Add your own tests

Click Add more tests in the Testing Agent tab, then choose how you want to add them:

• I’ll scan again and suggest more tests: the agent scans your app again for more flows.
• Write me some tests you want me to check: you describe the flows you want in plain words.

You can also describe your own test right in the prompt box:

“Test that the header nav opens Pricing, About, and Contact.”

“Add a task, then check that it appears in the task list.”

Write tests that work

Good tests are specific and end with something you can see on screen:

✓ “Submit the contact form and see a thank-you message.”
✓ “Add a product to the cart, open the cart, and reach the checkout page.”
✓ “Create a task and see it in the task list.”

Vague goals, or goals that happen off screen, don’t give you a useful result:

✗ “Make sure everything works.”
✗ “Check the backend.”
✗ “Test all edge cases.”
✗ “Check that the contact form email arrives.”

Each test has a 10-minute limit. Split long flows into steps, like “Booking step 1: pick a date” and “Booking step 2: confirm details.”

Running Tests

After you pick your flows, click Run tests. Your tests run in a real browser, and you can watch every step live as the agent clicks, types, and navigates through your app.

Every test:

• Runs on the preview of your app in the Studio, not your published site
• Uses a desktop-size screen
• Has a 10-minute limit
• Uses made-up test data, like [email protected], never your real details

You can run one set of tests per project at a time. To start a new run, wait for the current one to finish or click Stop run.

Testing Agent saves your tests with your project, so you can open the Testing Agent tab and run them again anytime.

Running tests from the tab

Once your project has tests, run them from the Testing Agent tab:

• Click Run all tests to run every test. Tick some tests first and the button changes to Run selected, so you run only those.
• Click a test’s own button to run just that test: Run test if it hasn’t run yet, Run again if it passed, or Run test again if it failed.
• While a test runs, click Watch test to see it live.
• Click Stop run to stop the current run.

To find a test, use search or filter by status. Switch between Table view and Replay view to see your tests the way you prefer.

Understanding Results

Each test ends with one of two results, plus the reason in plain words:

Result	Meaning
Passed	The agent completed the flow and saw the expected outcome.
Failed	The flow didn’t work as expected, or the agent couldn’t complete the test.

A failure tells you exactly what went wrong, for example:

“Both CTA buttons point to a broken link and return a 404 page instead of opening About us.”

When Testing Agent can’t test something, like checking an email inbox, it marks the test as Failed and explains why. Read the reason before you fix anything.

Before a test has a result, it shows Not tested yet (new or edited tests) or Testing… (while it runs). The summary at the top of the tab counts your Failed, Passed and Not tested yet tests.

The tester is an AI agent, so the same test can occasionally pass on one run and fail on the next. If a result looks off, run the test again.

Results belong to the version you tested

A Passed result means the flow worked on the version you tested. Every change you make, including changes from Fix with AI, creates a new version, and even a small change can break a flow that passed before.

Run your tests again after any change that touches a page or flow you tested.

Fix with AI

When a test fails, click Fix with AI on its row in the Testing Agent tab, next to Run test again. The agent picks up the failed test in the chat, fixes the issue, and then you choose:

• Run test again to check that the fix worked
• Skip to move on to the next test

Fix with AI uses credits like any other prompt.

Scan History

Scan history sits next to Findings in the Testing Agent tab. Findings holds your tests, and Scan history holds your past runs, grouped by version.

Each run shows how many tests passed and failed, with a video recording. Open a run to see the result of each test and rewatch exactly what the agent did and saw. Testing Agent keeps recordings for a limited time after each run.

Click Expand all to open every run, and Load more to see older ones.

Publishing with Failed Tests

Failed tests never block publishing. Treat them as a heads-up, fix what matters to you, and publish whenever you’re ready. See publishing your site for details.

Credits and Plans

Testing Agent is available on all plans, with no limit on how often you run it. Running tests uses credits: about 12 to 15 tests use 1 credit, so a full check of 6 tests uses roughly half a credit. Fix with AI uses credits like any other prompt.

When to Use Testing Agent

Testing Agent works best for:

• A final check before you publish or re-publish
• Catching what broke after a big change, like “did the redesign break my nav?”
• Links and navigation: CTAs, menus, footer links, and buttons that go to the right page
• Forms in your app: fields accept input, submit works, and your app shows a confirmation
• Interactive pieces: carousels, tabs, modals, accordions, filters, and search
• Flows that create their own data: add a task, then see it in the task list
• The path to payment: add to cart, open the cart, and reach the checkout page
• Status on screen: submit a request and see the status change to Pending
• Multi-step flows that finish within 10 minutes, like a booking or onboarding wizard

What Testing Agent Can’t Do

Testing Agent sees only what happens in the browser on your Studio preview. It can’t:

• Confirm emails or messages arrived. It submits your contact form, but it can’t check an inbox, Slack, or SMS.
• Test login, signup, logout, or password reset. Testing Agent leaves these flows out of your tests.
• Check how your layout works on phones or other screen sizes. Every test uses a desktop-size screen.
• Check your published site or custom domain. Every test runs on the Studio preview.
• Test flows that need existing data, a second user, or an admin role.
• Finish flows longer than 10 minutes. Split them into shorter tests.
• Give identical results every time. The same test can occasionally pass once and fail once.

When to Use Something Else

Testing Agent checks the flows you pick. For other needs, use a different approach:

Need	Better approach
Tests that give the same result every time and block releases	Playwright or Cypress end-to-end tests
Business logic or API contracts	Unit or integration tests
Security	Security Scanner
Performance and SEO	Lighthouse audit, and SEO & Discoverability for SEO
Accessibility	Accessibility Scanner
Design, content, or visual judgment	Your own review in the preview
Login, payments, emails, or anything with real-world effects	Your own check by hand
Monitoring your live site after publishing	A dedicated monitoring tool

FAQs

How many credits do my tests use?

About 12 to 15 tests use 1 credit, so a full check of 6 tests uses roughly half a credit. Fix with AI uses credits like any other prompt.

Can I use Testing Agent on my plan?

Yes. Testing Agent is available on all plans, with no limit on how often you run it.

Do I need to write my own tests?

No. The agent suggests up to 6 flows from your app, and you can ask for more, up to 20. You can also describe your own tests in the prompt box.

Can I edit a suggested test?

Yes. In the Testing Agent tab, open the test’s More actions menu, choose Edit test, and tell the agent what to change. The test shows Not tested yet until you run it again.

How do I run my tests again?

Open the Testing Agent tab and click Run all tests. To run only some tests, tick them and click Run selected. To run a single test, click Run again on its row, or Run test again if it failed.

Does a failed test stop me from publishing?

No. Results are a heads-up only. You can always publish.

Does Testing Agent check my published site?

No. It tests the preview in the Studio, before you publish.

Does it check that my contact form email arrives?

No. It checks that the form submits and your app confirms it. It can’t see the inbox.

Can I test my login or signup flow?

No. Testing Agent leaves login, signup, logout and password reset out of your tests.

Can I check how my app works on a phone?

No. Tests run on a desktop-size screen.

I changed my app after testing. Are my results still valid?

No, not automatically. Every change creates a new version, and a flow that passed before can break. If the change touched anything you tested, run those tests again.

Why did the same test pass once and fail once?

The tester is an AI agent, so results can occasionally vary between runs. Run the test again and read the reason for any failure.

Where do I see past results?

In the Testing Agent tab, under Scan history. Each run has a video recording, which Testing Agent keeps for a limited time.

Can I schedule test runs?

No. You run tests yourself from the Studio whenever you want.