Roadmap

Identify current limits and see which additions are proposed for Retest.

Check this page before planning a flow that needs a capability Retest does not yet support. It lists current limits and proposed additions; use the platform and feature guides for working examples.

The proposals have no approved dates or release order. They describe possible additions, so keep tests on the APIs available today.

Current limits Link to Current limits

  • Test structure. There are no automatic retries, flaky-result status, watch mode or custom fixtures with test.extend.
  • Parallel work. Files run on parallel workers, while tests within one file run in sequence.
  • Locators. There is no filter(), and(), or(), getByAltText() or getByTitle(). Closed shadow roots are inaccessible.
  • Firefox and WebKit. Frames, uploads, downloads, dialogs, popups, drag and drop and mocked responses are unsupported.
  • Firefox diagnostics. Network requests are captured, but console messages are unavailable.
  • Screenshot comparisons. There is no full-page capture or masking. Native app screens and Electron windows cannot be compared.
  • Platforms. Android apps, Safari, Windows and real phones or tablets are unsupported.
  • Secrets in images. Retest does not redact screenshots, so a secret displayed by the app can appear in captured pixels.
  • AI checks. Use test.evaluate for individual checks. Dataset evals with test.eval and the proposed toMeet matcher are unavailable.
  • Coding agents. There is no MCP server or installable skill.

These limits do not apply equally to every engine. For example, Chrome and Chromium support frames, and open shadow roots are searchable, with finder restrictions on Firefox and WebKit. Check Browsers and the relevant guide before writing a flow that needs them.

Proposed work Link to Proposed work

These additions are proposals. See the Changelog for capabilities already delivered in 0.2.0, including files, frames, dialogs and screenshot comparisons.

  • An MCP adapter would let coding agents use Retest's session operations through a tool server.
  • Retries would preserve a late pass as flaky, alongside custom fixtures and a watch mode that keeps the browser ready.
  • Broader Playwright compatibility would cover more existing suites, configuration and reporter integrations.
  • Android support would begin with device Chrome and native apps, with verification on emulators and real hardware.
  • Safari and further iOS browser support would extend testing beyond the current WebKit build and native simulator apps.
  • Evals would measure judges against labelled datasets and use explicit gates to decide whether a run passes.
  • Published benchmarks would compare measured runs and include cases where Retest is slower.
  • Windows support still needs a scope decision before implementation.

If your flow needs an unavailable operation, keep that limit visible when planning your tests. The Test API lists the methods you can use today, while Coming from Playwright explains the supported migration path.

Next steps Link to Next steps