Skip to main content
Glama
chrome-automation-mcp

chrome-automation-mcp

by JackZhao98

Chrome Automation MCP

中文版文档:docs/README-zh.md

A Model Context Protocol (MCP) server for browser automation using Playwright to control Chrome browsers.

npm version License: MIT

Installation

npm install -g chrome-automation-mcp@1.2.0

Requirements:

  • Node.js 18.0.0 or higher

  • Google Chrome browser

Related MCP server: Puppeteer MCP Server

MCP Configuration

Claude Desktop Setup

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "chrome-automation": {
      "command": "chrome-automation-mcp-full"
    }
  }
}

Lite Mode (Essential Tools Only)

{
  "mcpServers": {
    "chrome-automation": {
      "command": "chrome-automation-mcp"
    }
  }
}

Available Tools

🚀 Browser Management

  • launch_browser - Launch Chrome browser with session management

  • connect_browser - Connect to existing Chrome instance

  • close_browser - Close browser connection

📍 Navigation & Interaction

  • navigate_to - Navigate to URL

  • click - Click on elements with smart visibility detection

  • type_text - Type text into input fields

  • scroll - Scroll page to find hidden elements

  • wait_for - Wait for elements (auto-switches to new tabs)

📊 Information Gathering

  • read_text - Read text content from page/elements

  • get_elements - Get element information and attributes

  • screenshot - Take page screenshots

  • get_page_info - Get current page information

🖥️ Tab Management

  • switch_to_tab - Switch between tabs

  • get_tabs - Get information about all open tabs

💻 Code Execution

  • run_script - Execute external JavaScript files with browser access

  • evaluate - Execute JavaScript in browser context

  • set_storage - Set browser storage (cookies, localStorage, etc.)

⚙️ Session Management

  • list_sessions - List active browser sessions

  • press_key - Press keyboard keys with modifiers

  • go_back - Navigate back in browser history

Quick Start

  1. Launch Browser

{"tool": "launch_browser", "arguments": {}}
  1. Navigate to Website

{"tool": "navigate_to", "arguments": {"url": "https://google.com"}}
  1. Take Screenshot

{"tool": "screenshot", "arguments": {"fullPage": true}}

Script Development

Create custom automation scripts:

// my-automation-script.js
const searchQuery = args.query || 'MCP servers';

// Navigate to Google
await page.goto('https://google.com');

// Search
await page.fill('input[name="q"]', searchQuery);
await page.press('input[name="q"]', 'Enter');

// Wait for results
await page.waitForSelector('h3');

// Get all results
const results = await page.$$eval('h3', els => 
  els.map(el => el.textContent)
);

return { 
  query: searchQuery,
  searchResults: results,
  count: results.length
};

Use the script:

{
  "tool": "run_script", 
  "arguments": {
    "scriptPath": "./my-automation-script.js",
    "args": {"query": "playwright automation"}
  }
}

Available in scripts:

  • browser - Playwright browser instance

  • page - Current page object

  • args - Passed arguments

Lite Mode Features

Lite mode includes only essential tools for basic automation:

  • Browser management (launch_browser, close_browser)

  • Script execution (run_script)

  • Storage management (set_storage)

Perfect for lightweight integrations and custom script-based workflows.


Built with ❤️ for the MCP ecosystem

Available Tools

10 tools
cleanup_sessionsA

Clean up inactive/orphaned browser sessions and directories without affecting active sessions

Inputs: no named parameters.

close_all_browsersA

Gracefully close all active browser sessions (tries to close browsers properly first, then force kill if needed)

Inputs: force.

close_browserA

Gracefully close the browser connection and clean up session data

Inputs: sessionId.

get_loginA

🔐 INTERACTIVE LOGIN: Opens browser to specified URL, waits for manual login, then captures authentication data. A floating 'Finish Connect' button appears for you to click when login is complete. Button persists through page redirects during login flow.

Inputs: url (required), autoClose, saveToFile, waitMessage.

launch_browserB

Launch Chrome browser with session management. Each instance gets a unique session ID, port, and user data directory for isolation.

Inputs: debugPort.

list_sessionsB

List all active browser automation sessions

Inputs: no named parameters.

run_scriptB

📁 Execute a JavaScript file from local disk path or remote URL with access to browser automation. Use for complex automation scripts or reusable workflows. Must provide either scriptPath OR scriptUrl (not both).

Inputs: args, scriptUrl, sessionId, screenshot, +3 more.

run_script_backgroundB

📁 Execute a JavaScript file in the background with browser automation. Returns immediately with task info while script runs in background. Output is saved to files. Must provide either scriptPath OR scriptUrl (not both).

Inputs: args, scriptUrl, sessionId, screenshot, +5 more.

set_storageA

🔐 PREFERRED TOOL for setting login state and authentication data. Use this instead of evaluate() to set cookies, localStorage, and sessionStorage to maintain user login sessions across browser automation.

Inputs: sessionId (required), url, domain, cookies, +4 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 10 tools

    Disambiguation4/5

    Most tools have clearly distinct purposes (launch, navigate, script execution, storage, session management). Some potential overlap exists between cleanup_sessions and close_all_browsers, and get_login vs set_storage both touch auth data, but descriptions distinguish them adequately.

    Naming Consistency4/5

    Names follow a mostly consistent snake_case verb_noun pattern (launch_browser, run_script, close_browser, list_sessions). Minor deviations like navigate_to (verb only) and get_login (get is odd for an interactive capture tool) slightly break the pattern but remain readable.

    Tool Count5/5

    10 tools is well-scoped for a browser automation server, with each tool earning its place across launch, control, scripting, and cleanup concerns. No redundant or filler tools.

    Completeness4/5

    Covers the full lifecycle: launch, navigate, script execution, storage/auth, session listing, and cleanup/closing. Missing explicit page-interaction primitives (screenshot, click, get content) which agents must reach for scripts to accomplish, but core workflows are covered.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      B
      quality
      D
      maintenance
      A Model Context Protocol server that provides browser automation capabilities using Playwright. This server enables LLMs to interact with web pages, take screenshots, and execute JavaScript in a real browser environment.
      32
      13,875 npm
      5,682
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      A Model Context Protocol server that provides browser automation capabilities using Playwright, enabling LLMs to interact with web pages, take screenshots, and execute JavaScript in a real browser environment.
      19
      Apache 2.0
    • A
      license
      B
      quality
      D
      maintenance
      A Model Context Protocol server that provides browser automation capabilities using Playwright, enabling LLMs to interact with web pages, take screenshots, generate test code, scrape web content, and execute JavaScript in a real browser environment.
      32
      13,875 npm
      MIT
    • A
      license
      B
      quality
      D
      maintenance
      A Model Context Protocol server for web automation that enables browser control, element interaction, content extraction, and monitoring using Playwright.
      7
      20 npm
      ISC