Tencent has released BrowserSkill, a framework designed to connect AI agents to existing Chrome or Microsoft Edge browser sessions, allowing agents to operate within a user's authenticated environment. By utilizing a combination of a Rust-based CLI/daemon, a Chromium-based extension, and a specialized agent skill, BrowserSkill enables agents to perform complex web workflows such as filling forms, navigating internal sites, and capturing full-page screenshots while maintaining the user's current login state. The tool supports various shell-capable agents, including Cursor and Claude Code, and offers a dedicated plugin for DeepSeek Harness. Technical capabilities include advanced website debugging, which allows users to inspect HTTP request/response bodies, console messages, and performance metrics, as well as "remote mode" for pairing server-side agents with local browsers via authenticated WSS. For debugging, the system captures an operation timeline and provides request replay functionality. The architecture is implemented as a Rust and pnpm workspace, and the extension supports multiple languages, including English and several Asian and European languages. Users can manage session visibility through an Agent Window and can configure privacy settings to require human intervention for specific tasks like logins or verifications.
Read original
github-trending/typescript