What it does
This MCP server adds browser automation to Claude Code and other MCP clients using Playwright, WebDriver BiDi, and safaridriver. It exposes browser control directly as MCP-native tools, allowing you to navigate websites, interact with elements, fill forms, extract page content, and capture screenshots without leaving your AI assistant. Ideal for workflows that need live web interaction.
Who it's for
Developers using Claude Code to automate web tasks and interactions, QA engineers building browser-based test automation, and teams building AI-driven workflows that interact with live websites. Anyone who wants to script web interactions through an MCP protocol rather than managing standalone automation frameworks.
Common use cases
- Automate form filling and web navigation from Claude
- Extract and transform data from websites
- Capture screenshots for visual regression testing
- Run multi-step workflows involving web interactions
- Test designs across Chromium, Firefox, and WebKit browsers
Setup pitfalls
- 292 secrets found during scanning—conduct a thorough security audit before production deployment
- High-risk classification requires careful review of source code and dependencies
- CI pipeline currently failing; server may contain unstable code
- Requires filesystem write permissions for browser profiles, cache, and temporary files; requires network access to reach target websites
Playwrightbrowser binaries (Chromium, Firefox, WebKit) install separately during setup