Quickstart

This guide takes you from a product journey to a reviewed run. Start with one narrow journey, such as signup, checkout, invite acceptance or password reset. When its runs are reliable, add it to pull requests or a schedule.

Before you begin

1. Create a project

Open Projects and click New project. Name it after the product, such as "Marketplace web", and add its first environment: a name like "Staging" and the web URL. You can add Android and API targets to the environment later. See Projects and Environments.

2. Create a test case

In the project, open Test cases and click New test case. Give it a name, pick the platforms and the environment, then choose how to create it:

• Describe what to test: tell Autonomy the journey in plain words. It drafts outcome steps and asks a question when a choice would change the test.
• Write the steps: write each outcome yourself.
• Record it once: go through the journey once in a browser or on an Android emulator, and Autonomy drafts the steps from it.

Whichever you pick, you finish in the same steps editor. Write each step as an outcome someone could check, not as a click:

1. A new account exists for a fresh email.
2. The welcome email arrives and its link verifies the account.
3. The dashboard shows the starter plan as active.

To sign in, add a step for it and attach the account (Attach → Sign in as). To receive codes, attach a fresh inbox. Let the agent generate passwords and other test data instead of typing your own. See Authoring.

Before you save, Autonomy reads the steps in order and points out anything it would have to guess, such as which account signs in.

3. Run it

Click Run on the test case, or open New run, pick the environment, and click Start run. You can watch the run live. Under each step you see the agent's actions as they happen, ending with the Final check.

4. Review the result

Open the run and read What we checked: each step with its result and the actions behind it. If a step failed, its outcome line says what wasn't reached, and the replay, screenshots, network requests and emails show why. A step also fails when the agent couldn't show its outcome, and then it says why, for example that it ran out of time. Problems the agent saw along the way are listed under Issues found, apart from the result. See Runs and results.

5. Automate the useful path

When a test case passes reliably:

• Add it to a test plan and run the plan every night.
• Connect GitHub and test every pull request against its preview.
• Send failures to Slack and file issues as Linear or Jira tickets.

Keep the first automated plan small. Add broader journeys once the team trusts the results.