# WebTest AI CLI-first browser testing for teams that want AI assistance without handing pass/fail truth to a browser agent. WebTest AI turns human-readable specs into deterministic browser runs through a portable execution-driver layer. Reports preserve screenshots, traces, quality signals, and integration-friendly evidence. Model-backed workflows are optional and bounded to proposal flows such as discovery, recording cleanup, and reviewed healing. It is published under the AssertHive package scope: ```bash npm install --save-dev @asserthive/webtest-ai@beta npx webtest-ai --help ``` For a one-off invocation without installing it first: ```bash npx @asserthive/webtest-ai@beta --help ``` ## Why It Exists Modern product teams are moving toward agentic development, but browser quality still needs evidence, repeatability, and reviewable changes. WebTest AI is built around that line: - humans and agents can write simple specs - the runner converts those specs into known action types - deterministic browser execution owns pass/fail truth - reports preserve evidence that people and agents can inspect - healing and model workflows produce proposals instead of silently rewriting tests Compared with prompt-only browser agents, WebTest AI is designed to be a stable quality layer: agent-ready, deterministic, auditable, customizable, and safe to run locally, in CI, or through MCP-enabled agent workflows. ## Capabilities - Markdown suites with multiple tests per file - deterministic browser execution through the bundled Playwright driver - customizable execution-driver boundary for future or project-specific browser runtimes - accessibility-tree action surfaces for resilient role/name-oriented interaction - opt-in a11y and web-vitals quality checks - screenshots, traces, JSON reports, HTML reports, and CI summaries - guarded healing proposals backed by local UI inventory - recorder for turning browser interaction into spec drafts - model-backed discovery and cleanup when an explicit model profile is configured - MCP server for agents to read specs/reports/artifacts and run explicit debug tools ## Quick Start Install the beta package and the Chromium browser: ```bash npm install --save-dev @asserthive/webtest-ai@beta npx playwright install chromium ``` Create a local config when you want to override defaults: ```bash cp webtest-ai.config.example.json webtest-ai.config.json ``` Create a small suite at `specs/homepage.md`: ```bash mkdir -p specs ``` ```md --- suite: homepage baseUrl: https://example.com tags: [smoke] --- # Homepage Loads ## Steps 1. Open "/" ## Expected 1. Assert text "Example Domain" ``` Run it: ```bash npx webtest-ai run --suite specs/homepage.md ``` ## Commands - `webtest-ai run`: run a Markdown suite or directory of suites - `webtest-ai debug`: run with Playwright Inspector enabled - `webtest-ai record`: capture a browser flow into a Markdown spec draft - `webtest-ai heal`: review and approve verified UI inventory proposals - `webtest-ai discover`: generate bounded model-backed test-flow proposals - `webtest-ai report`: export reports or render CI summaries - `webtest-ai mcp`: start the MCP server over stdio Detailed command docs live at [the WebTest AI docs](https://asserthive.com/docs/webtest-ai), with a first-run guide at [the getting started guide](https://asserthive.com/docs/webtest-ai/guides/getting-started). The beta changelog lives at [the WebTest AI changelog](https://asserthive.com/docs/webtest-ai/changelog). ## Local-First Trust Model WebTest AI does not call a model by default. Deterministic runs, assertions, screenshots, quality checks, and reports work locally. Model-backed features require an explicit model profile and are limited to proposal-style workflows. Generated artifacts can contain application evidence. The built-in reports redact common headers, query params, emails, and bearer tokens, but raw Playwright traces may still contain data from the browser session. Treat `artifacts/` as test evidence and apply your normal retention policy. ## Config By default WebTest AI looks for: ```bash webtest-ai.config.json ``` Missing config files are allowed; built-in defaults are used. The example config is safe to commit: ```bash webtest-ai.config.example.json ``` Local runtime memory is written under `.webtest-ai/` by default and should not be committed. For the full configuration and API reference, see [the config and API reference](https://asserthive.com/docs/webtest-ai/config-and-api-reference). ## Beta Rollout WebTest AI is still in early beta. The npm package is published under the `beta` tag while the source repository remains private during early feedback. Try the CLI with: ```bash npm install --save-dev @asserthive/webtest-ai@beta npx playwright install chromium npx webtest-ai --help ``` The docs home for the package is [the WebTest AI docs](https://asserthive.com/docs/webtest-ai). ## License MIT © AssertHive