Is this an alternative to Playwright MCP?

Yes, while it uses Playwright technologies, it offers a Bash-based interface that is often faster and simpler for terminal-centric AI agents like Claude Code.

How does ref-based selection work?

The tool generates a snapshot where interactive elements are assigned simple IDs like @e1 or @e2; the agent then uses these IDs to click, type, or scrape.

Can I see the browser while it runs?

Yes, you can use the --headed flag to open a visible Chromium window, which is helpful for debugging automated flows.

What is agent-browser?

It is a CLI tool from Vercel designed for AI agents to interact with web pages using accessibility-based references instead of complex CSS selectors.

Does it support multiple browser sessions?

Yes, the --session flag allows you to run and maintain multiple independent browser contexts simultaneously without interference.

Agent Browser CLI Automation

Name: Agent Browser CLI Automation
Author: gvkhosla

bygvkhosla

•

浏览器自动化

Automates web browser interactions using a ref-based CLI optimized for AI agent workflows.

Agent Browser is a headless browser automation tool designed specifically for AI agents like Claude Code to interact with the web through standard Bash commands. By utilizing accessibility snapshots and a unique reference-based selection system (@e1, @e2), it allows agents to navigate pages, fill forms, extract data, and capture screenshots with high precision and minimal token overhead compared to traditional DOM manipulation. It serves as a lightweight, CLI-native alternative to Playwright MCP for developers who prefer bash-driven automation.

主要功能

0135 GitHub stars

02JSON output support for seamless data extraction and parsing

03Parallel browser sessions for managing multiple independent workflows

04Ref-based element selection using accessibility snapshots (@e1, @e2)

05Comprehensive interaction suite including click, fill, hover, and scroll

06Headless and headed browser modes for automation and debugging

使用场景

01Automated web scraping and data extraction from complex single-page applications

02Capturing full-page or viewport screenshots for visual verification and debugging

03Automated form submission and multi-step user flow testing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add gvkhosla/compound-engineering-pi agent-browser

For use in Claude.ai and ChatGPT

Download Skill