Getting Started
This guide takes you from nothing to a running test in two steps: create a project, then write your first test. By the end you will have an agent open your application in a real browser and report back on whether a flow works.
In a hurry? The Quick Start guide hands the whole setup to your coding agent and has you running in a couple of minutes.
1. Create a project
A project groups everything that belongs to one application: its tests, runs, and findings. It is just an organizational container, so name it after the thing you are testing. Most teams create one project per application, or one per environment such as staging and production.
In the sidebar, open Setup > Projects and click Create Project. Give it a name and an optional description, then save. That is all a project needs to get started; webhooks, failure reports, and context notes can be configured later from the project's detail page.
For the full list of project settings, see the Projects API reference.
2. Write your first test
A test is a re-runnable instruction written in plain language. You define it once and run it whenever you like; each run sends an agent to a real browser and links the result back to the test, so you build up a history of whether that flow keeps working.
Open Tests in the sidebar and click Create Test. The form has two parts.
- Prompt. The instruction the agent follows, including the URL to start from. For example: "Navigate to https://example.com, sign up with a new email, and verify the dashboard loads." Write it the way you would brief a person.
- Configuration. The device and browser the agent runs on. The defaults (a desktop browser on Chrome) are a fine starting point; advanced settings are tucked away until you need them.
Click Create test to save it, then run it. The agent opens the browser, works through your prompt, and reports back what happened along with screenshots. If it hits a problem, the run produces a finding describing what went wrong and how to reproduce it.