I built GhostTab: AI agents drive my real browser as text, in a hidden tab

Описание и применение

GhostTab позволяет ИИ-агентам управлять реальным браузером через текст в скрытой вкладке, экономя токены и избегая захвата экрана.

Задача

Screenshot-based browser agents were slow, expensive, and intrusive, taking over the screen.

Результат

In one day, an agent used GhostTab to submit ~40 job applications on LinkedIn Easy Apply and Greenhouse, improving speed and reducing token cost by 5–10×.

Рабочий процесс

• Start hidden browser with b bg on
• Navigate to URL with b go https://example.com
• Read page slice with b text 0 2000
• List clickable elements with b el
• Click element with b click 12
• Retrieve form fields with b form
• Fill form from JSON with b fillform @answers.json
• Perform OCR with b ocr total

Доказательства

Screenshot-based browser agents were slow and expensive for me, and they kept taking over my screen.

Источник

A screenshot costs roughly 1,500 tokens. A compact text view of the same page, with numbered elements, is usually a few hundred.

Источник

This week my agent used it to fill and submit about 40 job applications (LinkedIn Easy Apply and Greenhouse) in one day.

Источник

macOS only for now (AppleScript plus an optional extension bridge for background real input). Free and open source (MIT).

Источник