Automated web UI tests your whole team can read
Write browser tests in plain English with Serenity BDD, Cucumber and Selenium 4. Describe each page once in YAML, use the ready-made steps, and get a report with a screenshot of every failure.
A UI test without Java
Most UI tests need no new code. Describe the page's elements in a YAML file, using the most stable locator for each one:
# locators/login.yaml: the page is called "Login" url: / check: Login button elements: Username: "[data-test='username']" Password: "[data-test='password']" Login button: "[data-test='login-button']"
Then write the scenario with the generic UI steps:
Scenario: A registered user can log in Given I open the "Login" page When I type "standard_user" into "Username" And I type "${config:ui.password}" into "Password" And I click "Login button" Then I should be on the "Products" page
The password comes from configuration, which can be an environment variable or a CI secret, so it never appears in the test.
Ready-made UI steps
| Step | What it does |
|---|---|
Given I open the "Login" page | Opens a page described in YAML |
Then I should be on the "Products" page | Waits for the page's URL and its check element |
When I click "Login button" | Also hover over, check, uncheck and clear |
When I type "Ada" into "First name" | Clears the field first |
When I select "Price (low to high)" from "Sort order" | Drop-downs, by visible text |
When I save the text of "Order number" as "orderId" | Reuse it later as ${orderId}, in UI or API steps |
Then "Error message" should have the text "..." | Also contains the text and has the value |
Then "Cart badge" should be visible | Also not visible, enabled, disabled and checked |
Then I should see 6 of "Product cards" | Counts matching elements |
For an element inside a repeated item, such as one product's button, a template locator lets you write When I click "Add to cart button" for "Sauce Labs Backpack".
Stable by design
- No sleeps. Assertions retry until they pass or the timeout expires, and Serenity waits for elements before using them. How the waits work.
- Parallel and isolated. Scenarios run concurrently, each in its own browser.
- No driver management. Selenium Manager downloads the right ChromeDriver for you.
- Checked before it runs. A feature check reports unknown steps, pages and elements without opening a browser, and suggests the closest valid step.
- Page objects when you need them. For complex flows, business-level steps such as
When I check out with the details "Ada" "Lovelace" "SW1A 1AA"are backed by a page object, with templates included.
From manual test case to UI test
Paste a manual test case into Claude Code, Copilot or Cursor, and the framework's AI workflow captures the real page's locators, writes the scenario using only known steps, checks it, and runs it until it passes. Read the full walkthrough in how to convert manual test cases into automated tests.
See your own test case automated
Paste a manual UI test case into the live demo and watch it become a running scenario.