Skip to main content
Your tests are YAML files in your repository, under your own version control and code review. The format is published.

​
Where your tests live

Momentic tests and modules are *.test.yaml and *.module.yaml files in your repository. For dashboard-managed web tests, momentic import pulls them down to disk with their referenced modules and recreates the folder hierarchy. Mobile tests use the separate momentic-mobile CLI; it has no cloud import command. Read the V2 format reference, the web command reference, and Steps for the published format and step references.

​
Migrating an existing test suite

Use a coding agent through the MCP server with the coding agent skills to write the YAML from an existing spec. This is the recommended way to port a suite. Momentic ships no automatic converter from Playwright, Cypress, or Selenium. Keep your existing suite and CI gate while you port one bounded flow. Preserve its setup, authentication, data, assertions, waits, and cleanup. Label the new test, lint it, and inspect the selection with momentic list --labels migrated (or momentic-mobile list --labels migrated for mobile) before running it. Compare both runners on the same build and starting state, including an expected failure. Review recovery and classification as well as the pass status. Retire an old spec only after its replacement preserves the coverage and CI failure behavior you require. Keep unsupported browser, device, fixture, or component coverage in its existing runner. Follow the guide for your source framework:
  • Playwright: fixtures, storage state, browser projects, and network checks.
  • Cypress: session validation, intercepts, command chains, and plugins.
  • Selenium and Appium: page objects, waits, capabilities, and device support.
  • Maestro: subflows, launch/reset behavior, selectors, and variables.
Reuse seed endpoints and setup logic in a JavaScript step with the documented Node sandbox globals. Helpers that depend on a framework’s driver, fixtures, or imported packages need adaptation or an external setup script. authLoad reads cookies and local storage from a Playwright storage-state file; keep the auth bootstrap and validate any IndexedDB-dependent authentication separately.

​
Migrate out of Momentic

Momentic ships no exporter. See How Momentic works for the execution model. The YAML files are already in your repository, and the format is published, so you do not need to extract test source from a vendor first. Enterprise customers can ask for a one-time eject to open-source Playwright at contract termination. Ask your account team. Some steps have no equivalent in an open-source framework:
  • An AI action takes a goal in plain English and decides the clicks and the types itself.
  • An AI check reads the DOM, the accessibility tree, and a screenshot to judge a semantic assertion.
  • The step cache replays a resolved step with no model call.
  • In-flow classification gives a failed run a category and a triage action before the CLI exits.
  • Auto-maintenance repairs a step that moved.
A port to Playwright must express each of these as explicit code, or drop it. Deterministic preset steps, such as navigating to a URL, waiting, or clicking a CSS selector, map onto a single call in a code-first framework. A preset step that resolves a target in plain English, evaluates an assertion, or extracts data with AI does not. Results already export in standard formats. Use --reporter junit, --reporter json, or --reporter playwright-json so your dashboards and CI gates do not depend on Momentic’s. See Results and reporting and JUnit outputs.

​
What stays behind

If you leave Momentic, dashboard run history, AI memory, and step caches stay behind. They are organization-scoped Momentic state, not test source. See AI memory and Step caching.

​
Reduce AI dependence

Write preset steps where the action is stable. Keep AI steps for the parts that move. The step cache replays a resolved locator with no model call while its stored signals still match the page. See Step caching and Steps.