* 戳上方蓝字“皮球不皮”关注我

你是否曾想过,让AI助手像真人一样操作浏览器?自动填写表单、抓取数据、执行测试,甚至处理复杂的网页交互?这一切不再是科幻场景,而是通过ExecuteAutomation开源的Playwright MCP服务器变成了现实。

项目介绍:AI与浏览器的桥梁

Playwright MCP服务器是一个基于Model Context Protocol(MCP)的开源项目,它在大型语言模型(LLM)和网页浏览器之间建立了一座智能桥梁。该项目由ExecuteAutomation团队开发,充分利用了微软Playwright这一强大的浏览器自动化框架。

传统的网页自动化需要编写复杂的脚本,而Playwright MCP服务器的革命性在于,它让AI模型能够直接理解和执行浏览器操作指令。无论是Claude、ChatGPT还是其他支持MCP协议的AI助手,都能通过这个服务器获得真实的网页操作能力。

图片

核心亮点功能

1. 全面的浏览器控制能力

该项目支持Chromium、Firefox和WebKit三大浏览器引擎,AI可以创建浏览器实例、管理多个标签页、控制导航行为。这意味着AI能够在不同浏览器环境下执行一致性测试或兼容性检查。

2. 智能页面交互

AI可以通过这个服务器执行点击、输入、滚动、悬停等丰富的交互操作。更令人惊喜的是,AI能够理解页面上下文,智能选择最合适的操作策略,比如等待元素加载、处理弹窗等。

3. 数据提取与内容分析

服务器提供了强大的内容提取功能,AI可以从网页中获取文本、属性、HTML结构等信息,并结合自身的理解能力进行深度分析,为决策提供数据支持。

4. 高级监控功能

除了基本操作,项目还支持网络请求监控、性能指标收集、截图和PDF生成等高级功能,让AI能够全面掌控网页行为。

安装与使用方法

环境要求

  • • Node.js 20.0及以上版本

  • • 支持MCP协议的AI客户端(如Claude Desktop)

快速安装

<span data-cacheurl="" data-remoteid="" data-lazy-bgimg="https://mmbiz.qpic.cn/mmbiz_svg/K6CEv0Hv9DfqL3KtlSNbG8Vibx8PBZ9y7byFicLLqwWh29PC8j9qibObwVbOsRWNbvOicWmdjXdGhubeuVUKnvRacO6ibWMhRJQo6/640?wx_fmt=svg&amp;from=appmsg" data-fail="0"></span><code><span leaf="">npm install -g @executeautomation/playwright-mcp-server</span></code>

配置Claude Desktop

在Claude Desktop的配置文件(claude_desktop_config.json)中添加:

<span data-cacheurl="" data-remoteid="" data-lazy-bgimg="https://mmbiz.qpic.cn/mmbiz_svg/K6CEv0Hv9DfqL3KtlSNbG8Vibx8PBZ9y7byFicLLqwWh29PC8j9qibObwVbOsRWNbvOicWmdjXdGhubeuVUKnvRacO6ibWMhRJQo6/640?wx_fmt=svg&amp;from=appmsg" data-fail="0"></span><code><span leaf="">{</span><span leaf=""><br></span><span leaf="">&nbsp;&nbsp;</span><span><span leaf="">"mcpServers"</span></span><span leaf="">:</span><span leaf="">&nbsp;</span><span leaf="">{</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp;&nbsp;</span><span><span leaf="">"playwright"</span></span><span leaf="">:</span><span leaf="">&nbsp;</span><span leaf="">{</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">"command"</span></span><span leaf="">:</span><span leaf="">&nbsp;</span><span><span leaf="">"npx"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">"args"</span></span><span leaf="">:</span><span leaf="">&nbsp;</span><span leaf="">[</span><span><span leaf="">"-y"</span></span><span leaf="">,</span><span leaf="">&nbsp;</span><span><span leaf="">"@executeautomation/playwright-mcp-server"</span></span><span leaf="">]</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp;&nbsp;</span><span leaf="">}</span><span leaf=""><br></span><span leaf="">&nbsp;&nbsp;</span><span leaf="">}</span><span leaf=""><br></span><span leaf="">}</span></code>

Windows用户可能需要稍作调整:

<span data-cacheurl="" data-remoteid="" data-lazy-bgimg="https://mmbiz.qpic.cn/mmbiz_svg/K6CEv0Hv9DfqL3KtlSNbG8Vibx8PBZ9y7byFicLLqwWh29PC8j9qibObwVbOsRWNbvOicWmdjXdGhubeuVUKnvRacO6ibWMhRJQo6/640?wx_fmt=svg&amp;from=appmsg" data-fail="0"></span><code><span leaf="">{</span><span leaf=""><br></span><span leaf="">&nbsp;&nbsp;</span><span><span leaf="">"mcpServers"</span></span><span leaf="">:</span><span leaf="">{</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp;&nbsp;</span><span><span leaf="">"playwright"</span></span><span leaf="">:</span><span leaf="">{</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">"command"</span></span><span leaf="">:</span><span><span leaf="">"cmd"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">"args"</span></span><span leaf="">:</span><span leaf="">[</span><span><span leaf="">"/c"</span></span><span leaf="">,</span><span><span leaf="">"npx"</span></span><span leaf="">,</span><span><span leaf="">"-y"</span></span><span leaf="">,</span><span><span leaf="">"@executeautomation/playwright-mcp-server"</span></span><span leaf="">]</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp;&nbsp;</span><span leaf="">}</span><span leaf=""><br></span><span leaf="">}</span><span leaf=""><br></span><span leaf="">}</span></code>

代码演示:AI自动化实战

下面是一个典型的使用场景,展示AI如何通过MCP指令自动化网页操作:

<span data-cacheurl="" data-remoteid="" data-lazy-bgimg="https://mmbiz.qpic.cn/mmbiz_svg/K6CEv0Hv9DfqL3KtlSNbG8Vibx8PBZ9y7byFicLLqwWh29PC8j9qibObwVbOsRWNbvOicWmdjXdGhubeuVUKnvRacO6ibWMhRJQo6/640?wx_fmt=svg&amp;from=appmsg" data-fail="0"></span><code><span><span leaf="">// AI通过MCP工具执行网页自动化流程</span></span><span leaf=""><br></span><span><span leaf="">const</span></span><span leaf="">&nbsp;automationFlow = {</span><span leaf=""><br></span><span><span leaf="">steps</span></span><span leaf="">: [</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; {</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">action</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"navigate"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">url</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"https://example.com/login"</span></span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; },</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; {</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">action</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"fill"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">selector</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"#username"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">value</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"testuser"</span></span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; },</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; {</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">action</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"fill"</span></span><span leaf="">,&nbsp;</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">selector</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"#password"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">value</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"securepassword"</span></span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; },</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; {</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">action</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"click"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">selector</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"button[type='submit']"</span></span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; },</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; {</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">action</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"waitForNavigation"</span></span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; },</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; {</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">action</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"screenshot"</span></span><span leaf="">,</span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; &nbsp;&nbsp;</span><span><span leaf="">path</span></span><span leaf="">:&nbsp;</span><span><span leaf="">"dashboard.png"</span></span><span leaf=""><br></span><span leaf="">&nbsp; &nbsp; }</span><span leaf=""><br></span><span leaf="">&nbsp; ]</span><span leaf=""><br></span><span leaf="">};</span></code>

这个流程展示了AI如何自动完成登录操作并截图保存结果。在实际使用中,AI会根据页面实际情况动态调整选择器和等待策略。

应用场景详解

自动化测试

对于开发团队来说,这个工具可以大大提升测试效率。AI能够理解测试需求,自动生成和执行测试用例,甚至发现人工测试难以察觉的边缘情况。

数据采集与分析

市场研究人员可以使用AI自动收集竞争对手信息、价格数据、用户评论等,并进行实时分析,为决策提供数据支持。

日常办公自动化

从填写重复性表格到批量处理网页任务,AI可以接管大量繁琐的网页操作,让人类专注于更有创造性的工作。

无障碍测试

AI可以模拟不同用户的使用场景,自动检测网站的无障碍兼容性,帮助创建更加包容的网页体验。

优势对比:为何选择这个方案?

与传统自动化工具对比

相比Selenium、Cypress等传统自动化工具,Playwright MCP的最大优势在于"智能"。AI能够理解自然语言指令,适应页面变化,处理异常情况,而不需要编写复杂的条件判断逻辑。

与竞品MCP项目对比

虽然微软也有类似的playwright-mcp项目,但ExecuteAutomation的版本更加轻量化和易用。它专注于提供稳定可靠的核心功能,适合大多数应用场景。

开发体验优势

项目配置简单,文档清晰,社区活跃。开发者可以快速上手,遇到问题也能及时获得支持。

总结

Playwright MCP服务器代表了AI与自动化技术融合的重要方向。它不仅仅是一个工具,更是一种新的工作范式——人类描述意图,AI执行操作。

这个项目的意义在于降低了浏览器自动化的门槛。无论是技术专家还是业务人员,都能通过自然语言与AI协作完成复杂的网页任务。随着MCP协议的普及和AI能力的不断提升,我们可以预见未来会有更多类似的工具出现,进一步模糊人类与数字世界交互的边界。

关注公众号回复【源码】获取源码地址