Tencent Open-Sources 'BrowserSkill', Bridging AI Agents with Your Real Browser Sessions
Tencent has released BrowserSkill, an open-source bridge allowing AI agents to seamlessly share your actual browser sessions and tabs, eliminating redundant log
Image source: https://github.com/Tencent/BrowserSkill
Tencent has unveiled 'BrowserSkill,' a new open-source bridge tool designed to revolutionize the productivity of AI agents performing web tasks. While many existing automation tools provide agents with a blank browser context—requiring repetitive logins—BrowserSkill is designed to share your actual, authenticated browser session. This allows agents to initiate tasks immediately in a logged-in state.
This open-source release supports major AI agents and CLI workflows, including Cursor, Claude Code, and Codex, presenting a new automation paradigm for developers and productivity tool users.
BrowserSkill: Real-Time Sharing of Agent and Browser Sessions
The core advantage of BrowserSkill is that AI agents no longer need to generate new test accounts or fresh automation browser contexts. Instead, they share the browser session you are currently using, authenticated with your credentials.
- Persistent Authentication: Agents can utilize the state of websites you are already logged into (cookies, sessions, etc.), bypassing the need to perform MFA or complex login procedures again.
- Support for Major Agents: The tool provides official support for widely used AI agents and CLI environments, such as Cursor, Claude Code, and Codex, ensuring broad compatibility.
Intelligent 'Borrowing' for Seamless Automation
BrowserSkill also features an intelligent "tab borrowing" mechanism, designed to ensure agents perform their work without disrupting your workflow.
- Workflow Protection: Agents can "borrow" a specific tab to read or manipulate information required for a task and return it to you once the work is complete. This process maximizes agent efficiency without impeding your browsing workflow.
- Addressing Bot Detection: Because agents operate within a genuine, user-driven browser environment, they are relatively less impacted by the strict bot detection policies and security constraints frequently encountered by automated browsers.
Limitations and Caveats
While BrowserSkill provides a powerful automation environment, it is still subject to the browser’s own security settings and the strict bot detection policies of certain websites, which may occasionally restrict agent operations.