Can I see the browser while it is running?

Yes, you can use the --headed flag to launch a visible browser window, which is helpful for debugging interactions in real-time.

What makes Agent Browser different from standard Playwright?

Agent Browser is specifically optimized for AI agents using a 'snapshot + ref' pattern, which allows LLMs to interact with elements using simple IDs like @e1 rather than complex selectors.

How does the reference system work?

The snapshot command generates an accessibility tree where interactive elements are assigned short tags (e.g., @e1). You then use these tags in subsequent commands like 'click @e1' or 'fill @e1'.

Does it support parallel browsing?

Yes, using the --session flag, you can maintain multiple independent browser instances simultaneously for different tasks.

Agent Browser CLI

Name: Agent Browser CLI
Author: eyadsibai

byeyadsibai

Browser Automation

Automates web browser interactions using a reference-based accessibility tree optimized for AI agents.

About

Agent Browser is a headless browser automation CLI designed to bridge the gap between AI agents and the web. It leverages a unique accessibility snapshot pattern that allows Claude to interact with web elements via simplified references (like @e1 or @e2) instead of brittle CSS selectors. This skill is ideal for developers needing to scrape data, automate form submissions, or perform end-to-end testing directly from the command line, providing a reliable and bash-friendly workflow for complex web automation tasks.

Key Features

Multi-session support for managing parallel browser instances
Structured JSON output for easy data extraction and parsing
Headless and headed modes for automation and visual debugging
Reference-based element interaction using accessibility snapshots
Comprehensive interaction suite including click, fill, scroll, and hover
0 GitHub stars

Use Cases

Performing web scraping on JavaScript-heavy websites that require interaction
Automating complex multi-step form submissions and login flows
Generating screenshots and PDFs of web pages for automated reporting

About

Key Features

Multi-session support for managing parallel browser instances
Structured JSON output for easy data extraction and parsing
Headless and headed modes for automation and visual debugging
Reference-based element interaction using accessibility snapshots
Comprehensive interaction suite including click, fill, scroll, and hover
0 GitHub stars

Use Cases

Performing web scraping on JavaScript-heavy websites that require interaction
Automating complex multi-step form submissions and login flows
Generating screenshots and PDFs of web pages for automated reporting