Tencent has open-sourced BrowserSkill, a CLI tool and browser extension that enables AI agents to work within a user's real, logged-in browser session without disrupting active workflows.
BrowserSkill is an open-source bridge developed by Tencent that allows any shell-capable AI agent, including Cursor and Claude Code, to automate tasks directly within a user's active browser profile. Traditional browser automation agents rely on pristine, headless browser instances that struggle with logins, cookies, and bot detection; BrowserSkill sidesteps this by connecting a lightweight local CLI (`bsk`) to a browser extension that operates within designated tabs using existing sessions. It incorporates built-in human-in-the-loop controls to pause for CAPTCHAs, 2FA, or sensitive user approvals, providing a secure, pragmatic foundation for desktop web agent automation.
Headless browsers were the wrong abstraction for personal agentic workflows; tapping directly into authenticated user sessions solves the persistent login barrier while keeping safety controls where they belong.
- –Eliminates auth friction: Leveraging existing cookies and sessions bypasses the persistent wall of 2FA, SSO, and bot mitigation that cripples isolated headless browser environments.
- –Agent-agnostic integration: Operating over a standard CLI interface (`bsk`) makes it composable with any shell-enabled agent framework without bespoke SDK lock-in.
- –Pragmatic human-in-the-loop: Seamlessly pausing for CAPTCHAs and confirmation popups strikes the right balance between automation utility and operational supervision.
- –High-stakes blast radius: Granting agents access to logged-in sessions elevates the risk of unintended actions or prompt injection, making robust permission sandboxing and explicit user confirmation essential.
DISCOVERED
1h ago
2026-09-18
PUBLISHED
1h ago
2026-09-18
RELEVANCE