[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"pack-detail-browser-automation-zh":3,"seo:pack:browser-automation:zh":78},{"code":4,"message":5,"data":6},200,"操作成功",{"pack":7},{"slug":8,"icon":9,"tone":10,"status":11,"status_label":12,"title":13,"description":14,"items":15,"install_cmd":77},"browser-automation","🌐","#06B6D4","stable","稳定","浏览器自动化","Browser-Use \u002F Skyvern \u002F Stagehand 三巨头 + agent 真在用的 Chrome 扩展。选一个今天就能起爬虫。",[16,28,38,46,54,61,69],{"id":17,"uuid":18,"slug":19,"title":20,"description":21,"author_name":22,"view_count":23,"vote_count":24,"lang_type":25,"type":26,"type_label":27},709,"3d04e209-6c1a-4608-8e43-95b2cd7316d5","browser-use-ai-agent-browser-automation-3d04e209","Browser Use — AI Agent Browser Automation","Let AI agents control web browsers with natural language. Browser Use provides vision-based element detection, multi-tab support, and works with any LLM provider.","Browser Use",340,0,"en","mcp","MCP",{"id":29,"uuid":30,"slug":31,"title":32,"description":33,"author_name":34,"view_count":35,"vote_count":24,"lang_type":25,"type":36,"type_label":37},416,"6da285ea-0d45-4bf8-9f34-be39355dc7a7","skyvern-ai-visual-browser-automation-agent-6da285ea","Skyvern — AI Visual Browser Automation Agent","Automate any website using LLMs and computer vision. No selectors needed — works on sites never seen before. 21K+ stars.","Agent Toolkit",267,"skill","Skill",{"id":39,"uuid":40,"slug":41,"title":42,"description":43,"author_name":44,"view_count":45,"vote_count":24,"lang_type":25,"type":36,"type_label":37},688,"fd07b08f-5d31-404c-bd80-6e0587a6b712","stagehand-ai-powered-browser-automation-sdk-fd07b08f","Stagehand — AI-Powered Browser Automation SDK","TypeScript SDK that lets you automate browsers using natural language and visual understanding. AI sees the page like a human does. Built on Playwright. 10,000+ GitHub stars.","Browserbase",325,{"id":47,"uuid":48,"slug":49,"title":50,"description":51,"author_name":52,"view_count":53,"vote_count":24,"lang_type":25,"type":36,"type_label":37},415,"96b4872e-a870-40ad-b725-c35f20a99ca7","nanobrowser-ai-web-automation-chrome-extension-96b4872e","Nanobrowser — AI Web Automation Chrome Extension","Open-source Chrome extension with multi-agent AI for web automation. Free alternative to OpenAI Operator. 12K+ stars.","AI Open Source",301,{"id":55,"uuid":56,"slug":57,"title":58,"description":59,"author_name":22,"view_count":60,"vote_count":24,"lang_type":25,"type":36,"type_label":37},504,"badeea82-594a-4dbe-8cd0-9c3283235d1d","browser-use-web-ui-visual-ai-browser-automation-badeea82","Browser-Use Web UI — Visual AI Browser Automation","Gradio-based web interface for Browser-Use AI agent. Automate web browsing with visual feedback, persistent sessions, and HD recording. Supports 6+ LLM providers. 15,800+ stars, MIT.",330,{"id":62,"uuid":63,"slug":64,"title":65,"description":66,"author_name":67,"view_count":68,"vote_count":24,"lang_type":25,"type":36,"type_label":37},705,"8f2c0ae9-1327-481f-a519-d473751bdd76","crawlee-web-scraping-browser-automation-library-8f2c0ae9","Crawlee — Web Scraping and Browser Automation Library","Build reliable web scrapers in Node.js or Python. Crawlee handles proxy rotation, browser fingerprints, auto-scaling, and anti-bot bypassing out of the box.","Apify",272,{"id":70,"uuid":71,"slug":72,"title":73,"description":74,"author_name":75,"view_count":76,"vote_count":24,"lang_type":25,"type":26,"type_label":27},477,"17f46368-d971-4227-ab65-f5f183e7dedc","chrome-devtools-mcp-browser-debugging-ai-agents-17f46368","Chrome DevTools MCP — Browser Debugging for AI Agents","Give your AI coding agent full access to Chrome DevTools for browser automation, debugging, and performance analysis. Works with Claude, Cursor, Copilot, and 15+ AI tools.","MCP Hub",312,"tokrepo install pack\u002Fbrowser-automation",{"pageType":79,"pageKey":8,"locale":80,"title":81,"metaDescription":82,"h1":13,"tldr":83,"bodyMarkdown":84,"faq":85,"schema":101,"internalLinks":111,"citations":124,"wordCount":137,"generatedAt":138},"pack","zh","浏览器自动化：Browser-Use \u002F Skyvern \u002F Stagehand 七件套","Agent 真在生产用的七个浏览器自动化资产：Browser-Use \u002F Skyvern \u002F Stagehand 三巨头 + Chrome MCP 弹窗补丁 + 三个爬虫模板。TokRepo 一条命令装齐。","七个真在生产里跑的浏览器自动化资产 —— Browser-Use \u002F Skyvern \u002F Stagehand + Chrome MCP 弹窗补丁 + 三个爬虫模板。TokRepo 一条命令装齐。","## 这个 pack 装了什么\n\n这个 pack 收齐了**七个浏览器自动化资产**，是从公开 agent 配置、Twitter 截图反复反向工程出来的常客组合。三个是头部开源框架，三个是基于这些框架封装的爬虫模板，一个是多 agent 共用 Chrome 时不可或缺的运维 MCP 补丁。\n\n| # | 资产 | 类型 | 干什么 |\n|---|---|---|---|\n| 1 | Browser-Use | Python 框架 | LLM 原生的 Playwright 封装，Agent 直接对话 |\n| 2 | Skyvern | Python 服务 | 视觉 + LLM 的浏览器自动化，独立 server |\n| 3 | Stagehand | TypeScript 框架 | Browserbase 出品的生产级 SDK |\n| 4 | Chrome MCP 弹窗补丁 | MCP server | 拦截 modal \u002F cookie 弹窗不挡 agent |\n| 5 | 无头爬虫模板 | 爬虫配方 | Skyvern + cron 起步 |\n| 6 | 登录墙爬虫 | 爬虫配方 | OAuth + 会话复用模式 |\n| 7 | Agent 用 web reader | MCP server | 给 Claude Code \u002F Cursor 的 Markdown 抓取 |\n\n## 为什么要装\n\n大部分「AI 爬虫」教程止步于「打开这个页面读文字」。真实浏览器自动化要处理登录墙、cookie 横幅、modal 弹窗、无限滚动，以及多 agent 抢同一个 Chrome 会话时互相打架的事实。这个 pack 选的就是专门解决这些问题的七个资产，不是那 200 个收藏多但没人维护的替代品。\n\n三个头部框架各有自己的位置：\n\n- **Browser-Use** 入门最容易。Python 原生，自带 Playwright 封装，LLM 用自然语言就能驱动。一次性自动化和原型阶段最好用。\n- **Skyvern** 跑成 server，用视觉 + LLM 一起做。页面结构在多次访问之间会变、选择器经常失效的场景最佳。比 Browser-Use 慢但更抗造。\n- **Stagehand** 是生产级选项。TypeScript 优先，Browserbase 团队出品，专为成千上万个 agent 会话设计。要做面向客户的产品，选这个。\n\nChrome MCP 弹窗补丁必带，因为多 agent 跑 Chrome 早晚会撞到一个 agent 的 modal 挡住另一个 agent 点击的 bug。这个 MCP server 拦截最常见的弹窗模式自动清掉。\n\n## 一条命令装齐\n\n```bash\n# 装整个 pack\ntokrepo install pack\u002Fbrowser-automation\n\n# 或者只挑你真要用的\ntokrepo install browser-use\ntokrepo install skyvern\ntokrepo install stagehand\n```\n\nTokRepo CLI 把每个资产装到正确位置 —— Claude Code subagent 进 `.claude\u002Fagents\u002F`，Cursor 规则进 `.cursor\u002Frules\u002F`，Codex CLI 进 AGENTS.md。按你的栈选子集就行。\n\n## 注意坑\n\n- **别让 Browser-Use 和 Stagehand 同时跑在同一个 Chrome 配置上**，会抢活动 tab。要用就分开 user-data-dir，或者用 Browserbase 远程会话。\n- **Skyvern 要 GPU 或便宜的视觉模型兜底**。默认配置每屏调一次 GPT-4o，长爬钱包烧得快。\n- **登录墙爬取会撞到 reCAPTCHA 服务条款**。爬自己的后台或明确允许自动化的站。这个 pack 自带「从真实浏览器导出会话」模式，不必让 agent 自己登。\n- **Cookie 横幅不全是 GDPR**，很多站用作反爬陷阱。弹窗补丁解决常见的，剩下要手动加。\n- **无头检测真的存在**。Cloudflare \u002F DataDome 会指纹识别。要绕过，用 Stagehand + Browserbase，他们自带隐身层。\n\n## 这个 pack 不够用的时候\n\n目标是大规模爬单个高价值站（招聘、房产、电商）时，专用爬虫如 Apify 或自己写 Playwright 集群成本能比 agent 方案低 5-10 倍。Agent 浏览器自动化最适合：\n\n- 事先不知道要哪些字段\n- 页面布局每次访问都不一样\n- 想把爬取嵌进更大的 agent 任务（调研、获客、QA）\n\n要纯吞吐就别上 LLM，自己写选择器。要灵活性、要扛住布局变更，这个 pack 赢。",[86,89,92,95,98],{"q":87,"a":88},"Browser-Use 免费吗？","免费。Browser-Use 是 MIT 许可，你只付 LLM token 钱（Claude \u002F GPT-4 \u002F 开源权重模型都能跑）。库本身在任何能跑 Python + Playwright 的环境都能跑。Skyvern 和 Stagehand 也都是开源 —— Stagehand 有一个 Browserbase 托管的付费 Chrome 选项。",{"q":90,"a":91},"Codex CLI 或 Cursor 能用吗，不一定 Claude Code 吧？","七个资产都是工具无关的 —— 是 Python \u002F TypeScript 库，不是 Claude Code 专属 subagent。TokRepo CLI 会装对应工具的接线（Claude Code 用 subagent，Cursor 用规则，Codex CLI 进 AGENTS.md）。Browser-Use 还内建 OpenAI \u002F Anthropic 切换，搭配你 CLI 用的模型即可。",{"q":93,"a":94},"Skyvern 跟 Browser-Use 比怎么选？","Skyvern 用视觉（截图 + LLM）找元素，所以布局变更也能扛 —— 选择器派的爬虫这种情况就崩。Browser-Use 是 Playwright 选择器驱动，更快更便宜但更脆。经验法则：用 Browser-Use 做原型，上生产看是要视觉（Skyvern）还是规模（Stagehand）。",{"q":96,"a":97},"跟 MCP 服务器全家桶 pack 有啥区别？","MCP 全家桶聚焦协议层连接器 —— 浏览器 \u002F 数据库 \u002F 文件系统 MCP server，任何 agent 都能接。浏览器自动化是有立场的：明确选了哪几个浏览器工具、怎么写爬虫。Chrome MCP 弹窗补丁两个 pack 都有，因为对任何碰真浏览器的 agent 都是承重墙。",{"q":99,"a":100},"什么时候*不*该上 agent 浏览器自动化？","吞吐比灵活性重要时。如果是每小时爬同 100 个商品页、布局从不变，直接写 Playwright 选择器 —— 成本立省 10 倍、还少一个不稳定的 LLM 依赖。Agent 层适合探索性任务或页面未知的场景。",{"@context":102,"@type":103,"name":104,"description":105,"numberOfItems":106,"publisher":107},"https:\u002F\u002Fschema.org","CollectionPage","Browser Automation","Browser-Use, Skyvern, Stagehand, and the Chrome extensions agents actually drive.",7,{"@type":108,"name":109,"url":110},"Organization","TokRepo","https:\u002F\u002Ftokrepo.com",[112,116,120],{"url":113,"anchor":114,"reason":115},"\u002Fzh\u002Fpacks\u002Fmcp-server-stack","MCP 服务器全家桶","浏览器 MCP server 与这个 pack 互补",{"url":117,"anchor":118,"reason":119},"\u002Fzh\u002Ftools\u002Fclaude-code","Claude Code","这些浏览器 agent 的主要宿主",{"url":121,"anchor":122,"reason":123},"\u002Fzh\u002Fpacks\u002Fagent-memory-layer","Agent 记忆层","跨会话记住爬取上下文",[125,129,133],{"claim":126,"source_name":127,"source_url":128},"Browser-Use is an open-source library that enables AI agents to control a browser via Playwright","browser-use\u002Fbrowser-use on GitHub","https:\u002F\u002Fgithub.com\u002Fbrowser-use\u002Fbrowser-use",{"claim":130,"source_name":131,"source_url":132},"Skyvern automates browser-based workflows using LLMs and computer vision","Skyvern-AI\u002Fskyvern on GitHub","https:\u002F\u002Fgithub.com\u002FSkyvern-AI\u002Fskyvern",{"claim":134,"source_name":135,"source_url":136},"Stagehand is a production-grade browser automation framework from Browserbase","browserbase\u002Fstagehand on GitHub","https:\u002F\u002Fgithub.com\u002Fbrowserbase\u002Fstagehand",507,"2026-05-02T15:00:00Z"]