Browser Automation That Doesn't Steal Your Tabs

What it does

BrowserSkill connects shell-capable AI agents to your existing browser through a local Rust CLI and browser extension, giving them a dedicated Agent Window that shares your login state without touching your normal tabs. If the agent needs a tab you already have open, it explicitly borrows it and returns it when finished, leaving the rest of your browser alone.

The interesting bit

The whole pipeline stays on your machine: the agent calls the bsk CLI, which routes tasks via local IPC and WebSocket to the extension. There is no cloud relay. When the agent hits a captcha, confirmation dialog, or fresh login screen, it can pause and ask you to take over, then resume once you have cleared the hurdle.

Key highlights

• Agent-agnostic: any shell-capable harness works, with no lock-in to a specific model or framework
• Reuses your actual browser profile and active login sessions instead of requiring separate test accounts
• Runs tasks in a separate Agent Window so your normal browsing stays visible and uninterrupted
• Built-in human-in-loop for captchas and other steps that still require a real person
• Supports macOS, Linux, and Windows; Chrome and Edge are officially supported

Caveats

• Firefox support is planned but not yet available
• Other Chromium-based browsers are expected to work but are not explicitly tested

Verdict

Worth a look if you want AI agents to interact with authenticated web apps without maintaining separate browser profiles or surrendering your own tabs. Skip it for now if you are a Firefox-only shop or need fully headless, non-interactive automation.