Some of the most important tests you have are the ones that check fine grained details and business logic, like a new-account restriction that expires after a week. Or a promo code that only works Tuesdays under a full moon.
When spelling out the nitty-gritty in a prompt takes longer than clicking through it, you can record the flow instead. As you click through the flow in the live browser, the agent watches what you do and then converts your steps into a Playwright or Appium test, complete with the same page-object code the agent writes for every other flow in your suite.
How it works
Press record in the agent's chat, or ask the agent to record a flow.
If authentication or test-data helpers are already in your workspace the agent will run those automatically — if not, the agent will create them for the future.
In the browser, click through the flow. Don't worry about doing it perfectly. The agent will tidy up as you go. Your typing becomes one step per field, the noise (corrections, scrolls, double clicks) disappears, and the agent adds the checks you implied, like “the cart shows 1 item.” The agent can reword, reorder, or remove steps, but it never invents a step for something that didn't happen in the browser.
Press Create flow, and the agent replays everything in a clean browser before it writes a line of code. If a step fails, the replay stops and tells you why, and you can record the missing part, let the agent fix it, or drop the step. Once the replay passes, the agent writes the flow with page objects, reusing the ones your suite already has. Recording covers web apps.
Curious how it works? Read this →