Tencent/BrowserSkill

What it solves

BrowserSkill allows AI agents to interact with your web browser without interrupting your active work. It solves the problem of agents needing separate test accounts or isolated environments by allowing them to reuse your existing login states and session data while keeping their activities in a separate, visible "Agent Window."

How it works

It acts as a local bridge between an agent harness (like Cursor or Claude Code) and the browser. The agent sends commands via the bsk CLI to a local daemon, which then communicates via WebSocket to a browser extension. This extension automates tasks in a dedicated Agent Window, though it can explicitly "borrow" tabs from your main browser session when requested.

Who it’s for

Developers and users of shell-capable AI agents (such as Cursor, Claude Code, Codex, OpenClaw, and DeepSeek Harness) who want their agents to perform web-based tasks using their own authenticated sessions.

Highlights

• Session Reuse: Uses your real login state so agents don't need separate accounts.
• Non-Intrusive: Tasks run in a separate window, allowing you to continue working uninterrupted.
• Universal Agent Support: Works with any agent that can execute shell commands via the bsk CLI.
• Human-in-the-Loop: Allows users to manually take over for captchas, logins, or confirmation dialogs.
• Broad Compatibility: Supports Chrome and Microsoft Edge on macOS, Linux, and Windows.