playwright-mcp
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChinesePlaywright MCP (Microsoft Browser Automation & Live DOM Interaction)
Playwright MCP(Microsoft 浏览器自动化与实时 DOM 交互)
Based on microsoft/playwright-mcp: official Microsoft Playwright Model Context Protocol (MCP) server providing structured browser automation through accessibility snapshots, bypassing the need for heavy screenshot parsing or vision-only models.
基于 microsoft/playwright-mcp:官方的 Microsoft Playwright Model Context Protocol (MCP) 服务器,通过可访问性快照提供结构化浏览器自动化,无需依赖大量截图解析或纯视觉模型。
🎯 When to Use
🎯 使用场景
- Interactive Browser Automation: Navigating web pages, clicking buttons, selecting options, typing into inputs, and submitting forms.
- Accessibility Snapshot Navigation: Inspecting web interfaces via structured ARIA/accessibility trees, drastically saving token context compared to full HTML dumps.
- End-to-End (E2E) Test Execution: Validating authentication flows, checkout flows, and user journey funnels.
- Screenshot & Console Capture: Taking targeted viewport screenshots and intercepting browser console logs for live debugging.
- 交互式浏览器自动化: 浏览网页、点击按钮、选择选项、向输入框中输入内容以及提交表单。
- 可访问性快照导航: 通过结构化的 ARIA/可访问性树检查 Web 界面,与输出完整 HTML 相比可大幅节省 Token 上下文。
- 端到端(E2E)测试执行: 验证身份验证流程、结账流程和用户旅程漏斗。
- 截图与控制台捕获: 截取指定视口或元素的屏幕截图,并拦截浏览器控制台日志以进行实时调试。
🚀 Setup & Registration
🚀 设置与注册
MCP Configuration
MCP 配置
Register Microsoft Playwright MCP in your agent's configuration ( or IDE settings):
.mcp.jsonjson
{
"mcpServers": {
"playwright": {
"command": "npx",
"args": ["-y", "@playwright/mcp@latest"]
}
}
}在 Agent 的配置( 或 IDE 设置)中注册 Microsoft Playwright MCP:
.mcp.jsonjson
{
"mcpServers": {
"playwright": {
"command": "npx",
"args": ["-y", "@playwright/mcp@latest"]
}
}
}🛠️ Key Capabilities & Tools
🛠️ 核心能力与工具
- : Navigates to any target URL and returns the parsed accessibility tree.
browser_navigate - : Clicks an element identified by its accessibility reference ID.
browser_click - : Types text into specified form inputs or content-editable elements.
browser_type - : Captures full-page or element-specific screenshots.
browser_screenshot - : Evaluates custom JavaScript in the browser execution context.
browser_evaluate
- :导航至任意目标 URL,并返回解析后的可访问性树。
browser_navigate - :点击通过可访问性引用 ID 标识的元素。
browser_click - :向指定的表单输入框或 contenteditable 元素输入文本。
browser_type - :捕获整页或指定元素的屏幕截图。
browser_screenshot - :在浏览器执行上下文中计算自定义 JavaScript。
browser_evaluate
⚡ Integration with REA & Security Workflows
⚡ 与 REA 及安全工作流集成
- Dynamic XSS & CSP Verification: Navigate to test endpoints with live attack payloads to verify if CSP actively blocks script execution or if DOM-based XSS triggers alert dialogs.
- Automated Login & Auth Flow Testing: Automate the multi-step login flow for web applications (Clerk, Supabase, OAuth) to test authenticated routes and cookie security flags.
- E2E Post-Fix Validation: After implementing a security patch or header reconfiguration, run Playwright MCP to verify that the website continues to function seamlessly for real users.
- 动态 XSS 与 CSP 验证: 使用实时攻击载荷导航至测试端点,验证 CSP 是否有效阻止脚本执行,或 DOM 型 XSS 是否触发 alert 对话框。
- 自动化登录与身份验证流程测试: 自动执行 Web 应用的 多步骤登录流程(Clerk、Supabase、OAuth),以测试受保护路由和 Cookie 安全标志。
- 修复后的 E2E 验证: 完成安全补丁或响应头重新配置后,运行 Playwright MCP,验证网站对真实用户而言仍可正常使用。