GitHub
Open Source

Control Your Browser
with AI & CLI

PAGENT is a Page Agent that acts on real web pages. YAML-based UI testing, Chrome extension bridge, and MCP support for AI assistants.

Get Started View on GitHub
# Install globally
npm install -g @devload/pagent

# Start browser bridge
pagent bridge start

# Run YAML scenarios
pagent run scenarios/*.yaml

Features

Everything you need for browser automation and testing

📝

YAML DSL

Write test scenarios in simple, readable YAML. No coding required for basic automation tasks.

🎭

Playwright Powered

Built on Playwright for fast, reliable cross-browser testing with automatic waiting and retries.

🔌

Chrome Extension

Control your real Chrome browser via WebSocket bridge. Perfect for testing with real extensions.

🤖

MCP Integration

Connect with Claude Desktop or Claude Code. Let AI assistants control your browser directly.

📊

Rich Artifacts

Collect screenshots, HTML, console logs, network requests, HAR files, and Playwright traces.

⚡

Parallel Execution

Run multiple scenarios concurrently with configurable workers. CI-friendly with JSON output.

Architecture

How PAGENT connects everything together

┌─────────────────────────────────────────────────────────┐
│                    Chrome Browser                        │
│  ┌─────────────────────────────────────────────────────┐│
│  │              PAGENT Extension                       ││
│  │             (WebSocket Client)                      ││
│  └──────────────────────┬──────────────────────────────┘│
└─────────────────────────┼───────────────────────────────┘
                          │ ws://localhost:9222
                          ▼
┌─────────────────────────────────────────────────────────┐
│                 PAGENT Bridge Server                     │
│                 (WebSocket Server)                       │
└─────────────────────────┬───────────────────────────────┘
                          │
          ┌───────────────┼───────────────┐
          ▼               ▼               ▼
    ┌──────────┐   ┌──────────┐   ┌──────────┐
    │   CLI    │   │   MCP    │   │  Your    │
    │ Commands │   │  Server  │   │  Code    │
    └──────────┘   └──────────┘   └──────────┘

MCP Integration

PAGENT works as an MCP server, allowing AI assistants like Claude to control your browser directly. Ask Claude to navigate, click, fill forms, and capture screenshots.

browser_navigate
browser_click
browser_fill
browser_screenshot
browser_get_dom
browser_new_tab
browser_list_tabs
browser_console_logs
# Claude Code (CLI) - One command setup
claude mcp add pagent -s user -- npx -y @devload/pagent

# Claude Desktop - Add to config
{
  "mcpServers": {
    "pagent": {
      "command": "npx",
      "args": ["-y", "@devload/pagent"],
      "env": {
        "BROWSER_BRIDGE_URL": "ws://localhost:9222"
      }
    }
  }
}