Skip to main content

Tencent's New Open-Source Tool Gives AI Agents a Hands-On Edge

Tencent has just dropped a new open-source tool that's turning heads in the AI community. Called BrowserSkill, it's designed to give AI agents a more hands-on way to handle everyday web tasks—without stepping on your toes while you're browsing.

Think about this: you're in the middle of an important online research, and you want an AI assistant to pull up some data from another site. Normally, you'd have to switch tabs, maybe log in again, and wait for the agent to figure things out. BrowserSkill changes that. It runs in its own separate window, so the AI can do its thing while you keep working uninterrupted. And here's the kicker: it can reuse the login sessions you already have in your browser. No more re-entering passwords or going through two-factor authentication every time you ask the AI to check something.

What really sets BrowserSkill apart is its compatibility. It's not just for Tencent's own agents—it works with a bunch of mainstream ones, like Claude Code and Codex. The secret sauce is a simple command-line tool called bsk. You just plug it in, and your agent can start browsing the web like a pro.

In real-world tests, developers put BrowserSkill through its paces. They had it search for trending tech topics on Zhihu and summarize the key points—done in minutes. They also had it scrape high-rated comments from social platforms and generate a sentiment profile. All of it ran smoothly and quickly.

The magic behind the scenes? BrowserSkill uses a CLI-driven approach with an extended intermediary architecture. That means the AI model doesn't have to directly parse complicated web elements. Instead, it gets a simplified view, which speeds things up and makes the whole process more stable.

Industry folks are already calling this a game-changer. It bridges the "last mile" between AI agents and the real web, giving developers and content creators a new way to handle multiple tasks at once. If you're curious, you can check out the project on GitHub: https://github.com/Tencent/BrowserSkill.

Image

Key Points

  • BrowserSkill is Tencent's new open-source browser automation tool for AI agents.
  • It runs in a separate window, so it doesn't interfere with your current browsing.
  • It reuses your browser's login sessions, eliminating the need for repeated authentication.
  • Compatible with Claude Code, Codex, and Tencent's Workbuddy via a simple bsk CLI.
  • Tested successfully on tasks like summarizing Zhihu topics and scraping social media comments.
  • Uses a CLI-driven architecture to avoid complex web element parsing, boosting speed and stability.