Overview
Convert Playwright, Cypress, Puppeteer, Selenium, Appium, Maestro, or Mabl suites into Autonomy test cases with one prompt to your coding agent.
Existing suites already encode the journeys you care about. The selectors, waits, and page objects around them are the part that rots. Migration keeps the first and drops the second: every test becomes an Autonomy test case whose steps describe intent, and the Agent resolves them against the live product at run time.
Two paths
Dashboard import (Playwright only). Open Test Cases → Import, choose Playwright spec, and upload a .spec.ts or .js file. Autonomy reads it as untrusted source text and drafts intent-level steps from it; nothing in the script is replayed. This is the fastest route for suites without page objects or custom fixtures. The same dialog imports an OpenAPI spec into API steps and an Autonomy Test Case export for round trips.
Coding agent over MCP (every framework). Cypress, Puppeteer, Selenium, Appium, Maestro, and Mabl have no direct importer, and Playwright suites with heavy fixtures convert better this way too.
The one-prompt method
Connect your coding agent to Autonomy over MCP (see Connect a Client), open the repository that holds the suite, and give the agent a prompt like the one on each framework page. The agent reads each test, lists the Test Cases it would create with autonomy_test_case_list first so existing coverage is reused, and creates the rest with autonomy_test_case_create.
What maps, and what does not
| Source construct | Autonomy |
|---|---|
| Navigate / visit / goto | Navigate step |
| Click / tap / press | Tap/Click with a described target ("the checkout button"), never a selector |
| Fill / type / sendKeys | Type. Credentials become agent values; config becomes {{env.VAR}} |
| Expect / assert / should | Assert on visible text, element, or URL |
| Network intercepts | Validate Network |
| Screenshot comparisons | Visual Compare with a Figma or screenshot reference |
waitFor*, sleeps, retries | Dropped. The Agent waits on the outcome the next step needs |
| Page objects, fixtures, custom commands | Dropped. Shared setup becomes Environment runtime values or Preconditions |
beforeEach login | A leading Type/Tap sequence; or an ordered journey when the login belongs to another platform |
Two rules keep migrated cases honest:
- Targets are described, not copied.
#btn-submit-2becomes "the submit button". A selector is an implementation detail the product is free to change. - Values under test stay literal. If a test checks that
$48.00appears, keep$48.00. If a test types a throwaway email, make it agent-chosen.