agent-browser

name	agent-browser
description	Browser automation for AI agents using agent-browser CLI. Token-efficient alternative to Playwright MCP with accessibility-first element selection.

Agent Browser

Token-efficient browser automation CLI for AI agents. ~82% fewer tokens than Playwright MCP.

Prerequisites

# Install globally (one-time)
npm install -g agent-browser

# Verify installation
agent-browser --version

# If not installed, the agent should run the install command above before proceeding.

Quick Reference

# Install check
agent-browser --version

# Core flow: open → snapshot → action
agent-browser open <url>
agent-browser snapshot          # accessibility tree (@e1, @e2, ...)
agent-browser click @e3
agent-browser fill @e5 "text"
agent-browser screenshot /tmp/page.png

Core Commands

Command	Description
open <url>	Open a page
snapshot	Return accessibility tree + element refs
click <@ref>	Click an element
fill <@ref> <text>	Fill an input field
type <@ref> <text>	Type text (preserves existing)
press <key>	Press key (Enter, Tab, Control+a)
screenshot [path]	Take screenshot
pdf <path>	Save as PDF
wait <sel|ms>	Wait for element or time
scroll <dir> [px]	Scroll (up/down/left/right)
close	Close browser

Element Selection

Accessibility tree @ref is the default:

agent-browser snapshot
# Output example:
# @e1 heading "Welcome"
# @e2 button "Login"
# @e3 textbox "Email"

agent-browser click @e2
agent-browser fill @e3 "[e-mail]"

CSS selector and XPath also supported:

agent-browser click "#submit-btn"
agent-browser click "xpath=//button[@type='submit']"

Find Elements

agent-browser find role button click         # Find by role and click
agent-browser find text "Submit" click       # Find by text
agent-browser find label "Email" fill "[e-mail]"
agent-browser find testid "login-btn" click

Get Info

agent-browser get text @e1       # Get text content
agent-browser get html @e1       # Get HTML
agent-browser get value @e3      # Get input value
agent-browser get url            # Current URL
agent-browser get title          # Page title
agent-browser get count "li"     # Element count

State Check

agent-browser is visible @e1
agent-browser is enabled @e2
agent-browser is checked @e3

Network

agent-browser network requests                    # List requests
agent-browser network route "*/api/*" --abort     # Block requests
agent-browser network route "*/api/*" --body '{}' # Mock response
agent-browser network unroute "*/api/*"

Browser Settings

agent-browser set viewport 1280 720
agent-browser set device "iPhone 15"
agent-browser set media dark
agent-browser set offline on

Workflow Patterns

1. Page Exploration

agent-browser open "http://localhost:3000"
agent-browser snapshot
# → Check @ref, then take action
agent-browser click @e5
agent-browser snapshot   # Verify state change

2. Form Filling

agent-browser open "http://localhost:3000/login"
agent-browser snapshot
agent-browser fill @e3 "[e-mail]"
agent-browser fill @e4 "password123"
agent-browser click @e5   # submit button
agent-browser wait 2000
agent-browser screenshot /tmp/after-login.png

3. Verification

agent-browser open "http://localhost:3000"
agent-browser wait "h1"
agent-browser get text "h1"
agent-browser is visible "#success-message"

When to Use

Scenario	Recommended
AI agent web browsing/verification	agent-browser (token efficient)
Complex E2E test suites	Playwright (richer API)
Advanced network interception	Playwright
Quick UI checks	agent-browser (simple CLI)