Hover

Hover works by leveraging the coding-agent CLI that you already run—Claude Code, Codex, or Gemini—to explore your web application through a real Chrome browser via Playwright MCP. The agent autonomously navigates pages, triggers interactions, and then converts each user flow into a plain @playwright/test file that you own. The generated tests are standard Playwright specifications, so they can be executed in CI without any AI runtime dependency.

The main advantage is bridging the gap between AI-driven exploration and maintainable test automation: you don't have to write test scripts manually; the agent discovers interaction paths and generates well-structured Playwright tests that can be version-controlled, reviewed, and run like any traditional test. Because everything happens locally, your browser data and application state stay private, which is appealing for teams with strict data policies.

However, the entry barrier is notable. You need to have a coding-agent CLI installed and configured (API keys, possible usage costs), and you must be comfortable with the respective toolchain.