How to use Chrome MCP: a simple guide

1 Introduction

mcp-chrome (also called Chrome MCP Server) is a Chrome extension–based MCP (Model Context Protocol) server. It lets MCP clients (e.g. AI agents, chatbots) connect to your regular Chrome/Chromium browser, use your existing login state, and control or inspect web pages.

Main GitHub Repo is here

2 Main Features

• Works with your real browser (no separate headless instance)
• Fully local (no remote server)
• Uses streamable HTTP as connection method (or stdio alternative)
• Semantic search across tabs (vector index) & content analysis
• Over 20 “tools” (APIs) to interact with browser: navigation, network, screenshots, DOM, bookmarks, history, etc.

3 Installation & Usage Guide

3.1 Requirements

• Node.js = 18.19.0 and npm or pnpm
• Chrome or Chromium browser

3.2 Installation Steps

1. Install the bridge globally:

   1	npm install -g mcp-chrome-bridge

   Or with pnpm (enabling pre/post scripts):

   1	pnpm config set enable-pre-post-scripts true && pnpm install -g mcp-chrome-Bridge

2. Register the bridge:

   1	mcp-chrome-bridge register

3. Download & install the Chrome extension: Go to the GitHub Releases of mcp-chrome and download the extension.

4. In Chrome, open chrome://extensions/, enable Developer mode, “Load unpacked” and select the extension folder, then open the extension’s popup.

3.3 Using It

• In the extension popup, click Connect to establish the bridge connection.

• The extension will show a port (default 12306) for the MCP server.

• In your MCP client config, add:

  1 2 3 4 5 6 7 8

  { "mcpServers": { "chrome-mcp-server": { "type": "streamableHttp", "url": "http://127.0.0.1:12306/mcp" } } }

  (Port 12306 is the default shown in extension popup)

• Alternatively, if your client only supports stdio mode, use the mcp-server-stdio.js shipped in the bridge and configure as a command/args in MCP client config.

4 Tools (APIs) in mcp-chrome

The following is a more complete list of tools exposed by mcp-chrome (v0.0.6)

4.1 Browser Management

• get_windows_and_tabs — list all browser windows & tabs
• chrome_navigate — navigate to URL, control viewport
• chrome_close_tabs — close specific tabs or windows
• chrome_go_back_or_forward — browser back / forward navigation
• chrome_inject_script — inject content scripts into pages
• chrome_send_command_to_inject_script — send commands to scripts you injected

4.2 Screenshots & Visual

• chrome_screenshot — capture screenshot, full page or element-level

4.3 Network Monitoring

• chrome_network_capture_start / chrome_network_capture_stop — use webRequest API to capture network activity
• chrome_network_debugger_start / chrome_network_debugger_stop — use Debugger API to capture deeper request/response data
• chrome_network_request — send custom HTTP requests from browser context

4.4 Content & Analysis

• search_tabs_content — semantic search across all open tab contents
• chrome_get_web_content — extract HTML / text content from a page
• chrome_get_interactive_elements — get clickable / interactive elements on page
• chrome_console — capture and retrieve console logs from browser tabs

4.5 Interaction (User Actions)

• chrome_click_element — click element via CSS selector
• chrome_fill_or_select — fill forms, select dropdowns, set values
• chrome_keyboard — simulate keyboard events or shortcuts

4.6 Data Management

• chrome_history — search browser history with time filters
• chrome_bookmark_search — search bookmarks by keyword
• chrome_bookmark_add — add a bookmark (with folder support)
• chrome_bookmark_delete — delete a bookmark

This set covers most of the common browser automation needs—navigation, content reading, network capture, user interactions, bookmarks, history, and more.

⚠️ Note: Because mcp-chrome is under active development, new tools might be added, or some may evolve. Always check the latest TOOLS documentation in the repo.