Claude Code evolves again

Claude Code evolves again! The built-in browser allows AI to "brush the web" independently. The well-known AI company Anthropic recently ushered in a major upgrade for its developer tool Claude Code, officially adding a built-in integrated browser window. This new feature gives Claude the ability to open, read, click, and type on external web pages directly within the application. Through this upgrade, developers can save the trouble of frequently switching applications and let AI efficiently retrieve key resources such as technical documents and tracking issue records directly within the tool. The built-in browser adopts a convenient tab management structure, and users can quickly wake it up with simple keyboard shortcuts. Taking both security and control into consideration, in order to prevent AI from unexpected overreaching behavior, Anthropic has deployed multiple strict security review mechanisms at the bottom of the technology. The classifier within the system will strictly screen all write operations on external websites to ensure that core control is always in the hands of the user. Without explicit authorization from the user, Claude will not be able to independently perform any operations to purchase goods, create new accounts or crack verification codes (CAPTCHA). Additionally, the browser always runs in a completely clean, isolated environment configuration and does not save any historical login credentials. Flexible organizational management: To address the security concerns of enterprise-level users, organizations can freely limit the range of external websites that AI can access by setting a whitelist. If an enterprise has higher-level compliance requirements, management may even choose to disable the browser tool entirely. For those users who want Claude to be able to perform operations in a personal logged-in session, the official recommendation is to use a dedicated Chrome browser extension. This set of combos not only demonstrates the rapid evolution of AI agent technology, but also builds a secure line of defense for the actual implementation of the tool. via AI News (author: AI Base)