WebTest AI Docs
WebTest AI is a CLI-first browser quality tool from AssertHive. It runs human-readable specs through deterministic browser automation and writes evidence that humans, CI systems, and agents can inspect.
V2 adds intent-driven testing for teams that maintain whitelabel, branded, or translated applications. One goal-based Markdown spec can run across configured targets while WebTest AI keeps browser execution, pass/fail truth, reporting, and safety checks deterministic.
V2 Highlights
- Bounded Goal Planner:
modelMode: intentturns## Goalinto a validated plan made only of supported browser actions. - Target matrix: define
baseUrl,locale,brand, and optionalintentAliasesso one suite can run across whitelabel or translated apps. Use the target matrix to select the brands and locales a spec should cover. - Semantic outcomes: use semantic
Assert outcomechecks when copy varies, while exactAssert textremains available for copy-sensitive checks. - CDP and Playwright parity: Chromium core-action workflows can run through bundled CDP or Playwright drivers; Playwright remains the multi-engine and richest-debug path.
- Auto Exploration:
webtest-ai discover --explorecan gather bounded browser evidence and propose flows from real pages. - UI Inventory: discovery and healing workflows can propose reviewed inventory updates instead of silently rewriting specs.
- Release evidence: reports include target metadata, generated intent plans, selected candidates, model call metadata, semantic evidence, screenshots, traces, and CI summaries.
Try V2 Locally
Run the bundled demo evidence pack:
npm run demo:v2
npm run demo:intent -- --headedThe demo runs the same intent specs across northwind-en, maison-fr, and mercado-es, covering account signup, login, product search, cart, discount, checkout, support, preview flows, Auto Exploration, and UI Inventory proposals.