AI News AI资讯 5h ago Updated 2h ago 更新于 2小时前 49

Claude Code now has a built-in browser that lets the AI read, click, and type on external websites Claude Code 现在内置浏览器,允许 AI 读取、点击和输入外部网站内容

Anthropic has integrated a built-in browser into Claude Code, enabling the AI to navigate, read, click, and type on external websites directly within the development environment. The feature utilizes a clean, isolated browser profile with no saved logins and employs classifiers to screen write actions, preventing unauthorized purchases, account creation, or CAPTCHA bypassing without explicit user consent. Organizations maintain strict control over the tool via allowlists for external site access Anthropic为Claude Code内置了浏览器功能,允许AI直接在应用内读取、点击和输入外部网站内容 该功能采用基于标签页的浏览器界面,通过快捷键启动,支持文档站点和问题追踪器等资源访问 引入多重安全机制,包括写入操作分类器筛查、禁止自动购买/注册/CAPTCHA绕过及无登录信息的干净配置文件 组织可通过白名单限制外部站点访问或完全禁用该工具,需登录会话的用户建议使用Chrome扩展

75
Hot 热度
65
Quality 质量
70
Impact 影响力

Analysis 深度分析

TL;DR

  • Anthropic has integrated a built-in browser into Claude Code, enabling the AI to navigate, read, click, and type on external websites directly within the development environment.
  • The feature utilizes a clean, isolated browser profile with no saved logins and employs classifiers to screen write actions, preventing unauthorized purchases, account creation, or CAPTCHA bypassing without explicit user consent.
  • Organizations maintain strict control over the tool via allowlists for external site access or the ability to disable browser tools entirely, while users requiring logged-in session interactions are directed to use a separate Chrome extension.

Why It Matters

This update significantly enhances the utility of AI coding assistants by bridging the gap between code generation and real-world documentation or API verification, allowing developers to fetch live context without leaving their IDE. It also highlights Anthropic's focus on safety and enterprise governance, providing concrete mechanisms to prevent autonomous agents from performing unintended or risky actions on the open web.

Technical Details

  • Integration & Interface: The browser functions as a tab-based interface within Claude Code, accessible via keyboard shortcuts, leveraging existing tool-use infrastructure adapted for web interaction.
  • Safety Mechanisms: Write actions on external sites are screened by classifiers; the system is hard-coded to refuse buying items, creating accounts, or bypassing CAPTCHAs unless the user explicitly consents.
  • Isolation Profile: The browser operates on a clean profile devoid of saved logins or cookies to prevent accidental data leakage or unauthorized session usage.
  • Enterprise Controls: Administrators can restrict access to specific domains using an allowlist or completely disable the browser tools, ensuring compliance with organizational security policies.

Industry Insight

  • Shift Toward Autonomous Agents: This move signals a broader industry trend toward AI agents capable of interacting with dynamic web environments, necessitating robust safety frameworks to manage autonomous browsing risks.
  • Security by Design: Developers should prioritize implementing granular permission controls and isolated execution environments when integrating AI tools that require external network access to mitigate liability and data exposure.
  • Hybrid Interaction Models: The distinction between the built-in secure browser and the Chrome extension for logged-in tasks suggests a future where AI tools will offer modular interaction modes tailored to different security and functionality requirements.

TL;DR

  • Anthropic为Claude Code内置了浏览器功能,允许AI直接在应用内读取、点击和输入外部网站内容
  • 该功能采用基于标签页的浏览器界面,通过快捷键启动,支持文档站点和问题追踪器等资源访问
  • 引入多重安全机制,包括写入操作分类器筛查、禁止自动购买/注册/CAPTCHA绕过及无登录信息的干净配置文件
  • 组织可通过白名单限制外部站点访问或完全禁用该工具,需登录会话的用户建议使用Chrome扩展

为什么值得看

这项更新标志着AI编程助手从本地代码环境向Web交互的重要延伸,使开发者能更高效地获取在线文档和调试信息。同时,其严格的安全设计为AI在受限企业环境中执行网络操作提供了可信赖的范式。

技术解析

  • 集成架构:浏览器作为Claude Code的内置组件运行,模拟标准标签页浏览器行为,通过键盘快捷键快速调出,无需切换上下文至外部浏览器。
  • 安全沙箱与权限控制:运行于无保存登录信息的干净配置文件中;内置分类器实时筛查所有写入动作;默认禁止涉及金融交易、账户创建或CAPTCHA破解的操作,除非获得用户明确同意。
  • 企业级管理策略:提供细粒度管控选项,管理员可通过配置白名单限制可访问的外部域名,或直接全局禁用浏览器工具以符合合规要求。
  • 替代方案指引:对于需要利用现有登录会话的场景,官方推荐改用独立的Chrome扩展,实现了内置工具与个性化扩展的功能分离。

行业启示

  • AI Agent安全性成为核心竞争力:Anthropic通过内置的安全筛查和权限隔离,展示了如何在赋予AI强大交互能力的同时保障企业数据与操作安全,这将成为未来AI工具落地的关键门槛。
  • 开发工作流的无缝融合:将Web浏览能力直接嵌入IDE/编程助手,减少了开发者在代码编辑器与浏览器之间切换的认知负荷,预示着AI辅助开发将进一步向全栈、全场景交互演进。
  • 灵活的分层部署策略:区分“内置安全模式”与“扩展个性化模式”,表明AI产品正在适应不同用户群体(如企业IT管控 vs 个人开发者自由探索)的差异化需求。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Claude Claude Agent Agent Code Generation 代码生成 Security 安全