Does it support JSON output?

Yes, adding the --json flag to snapshot commands returns structured data that is easy for Claude to parse and process programmatically.

Can I see the browser window while it runs?

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

Can I run multiple browser sessions at once?

Yes, the skill supports named sessions via the --session flag, allowing you to manage multiple independent browser instances simultaneously.

How does this compare to Playwright MCP?

While Playwright MCP uses tool-based integration, agent-browser is a CLI-first tool optimized for Bash-based workflows and quick, one-off automation tasks.

What is the benefit of ref-based selection?

Ref-based selection (@e1, @e2) simplifies interaction for AI agents by mapping interactive elements to short identifiers from accessibility snapshots, replacing complex CSS selectors.

Agent Browser Automation

Name: Agent Browser Automation
Author: 8b-is

by8b-is

ブラウザ自動化

Automates web browsing, form filling, and data extraction using the Vercel agent-browser CLI with ref-based selection.

概要

The agent-browser skill enables Claude Code to interact with the web directly through a powerful CLI interface. It uses a unique ref-based selection system derived from accessibility snapshots, making it significantly easier for AI agents to precisely click, type, and navigate pages without relying on complex or brittle CSS selectors. This skill is ideal for automating repetitive web tasks, performing headless testing, and extracting structured data from modern web applications using a workflow optimized specifically for LLMs.

主な機能

0 GitHub stars
High-resolution viewport and full-page screenshot capture
Parallel browser session management for multi-page workflows
JSON-formatted snapshots for structured data extraction and parsing
Interactive form automation including text entry, checkboxes, and dropdowns
Ref-based element selection using accessibility snapshots (@e1, @e2)

ユースケース

Generating visual snapshots or PDF reports of web pages for documentation
Scraping dynamic content from Single Page Applications (SPAs)
Automating complex end-to-end user flows like login and account setup

概要

主な機能

0 GitHub stars
High-resolution viewport and full-page screenshot capture
Parallel browser session management for multi-page workflows
JSON-formatted snapshots for structured data extraction and parsing
Interactive form automation including text entry, checkboxes, and dropdowns
Ref-based element selection using accessibility snapshots (@e1, @e2)

ユースケース

Generating visual snapshots or PDF reports of web pages for documentation
Scraping dynamic content from Single Page Applications (SPAs)
Automating complex end-to-end user flows like login and account setup