Chrome MCP Server turns your existing Chrome browser into a local, model-driven automation surface. By leveraging the mcp-chrome-bridge, it exposes your browser to MCP clients and enables a range of tools for browser control, content analysis, and semantic search. The bridge runs locally and supports a streamable HTTP connection, allowing clients to send commands and receive results in real time. Once the bridge is installed and running, you can connect any MCP client to the Chrome browser environment and drive tasks like tab management, navigation, content extraction, and script injection through natural language or structured commands. The available browser tools include managing windows and tabs, navigating to URLs, injecting scripts, performing in-page commands, and interacting with the browser’s native APIs, all while preserving your existing login states and configurations.

Prerequisites:

• Node.js (recommended: v20.x or newer) and npm/pnpm
• Chrome or Chromium browser installed

Installation steps:

1. Install the Chrome MCP Bridge globally via npm or pnpm (as described in the README):

npm install -g mcp-chrome-bridge

or using pnpm (recommended for some setups):

pnpm config set enable-pre-post-scripts true pnpm install -g mcp-chrome-bridge

2. Optionally, verify the installation:

mcp-chrome-bridge --version

3. Load the Chrome extension and connect to MCP:

• Open Chrome, navigate to chrome://extensions, enable Developer mode, and load the unpacked extension folder downloaded from GitHub.
• Use the extension to connect and obtain the MCP configuration (mcpServers) to use in your MCP client configuration.

4. Start the MCP bridge via the configured command in your MCP client (as shown in the mcp_config example):

Example run (via MCP client using npx):

npx -y mcp-chrome-bridge

Practical tip: If you rely on pnpm, you may need to register or adapt commands as shown in the README (e.g., using the manual registration flow if postinstall scripts are restricted).

Tips and common considerations:

• The bridge enables streaming HTTP control for low-latency browser automation. Prefer using the streamable HTTP connection in your MCP client configuration.
• Ensure the Chrome extension is loaded and connected before issuing MCP commands; the extension provides the native browser API access required by the tools.
• Environment variables are typically not required for basic operation, but advanced setups may benefit from configuring paths to user data directories or specific browser profiles if you need isolation.
• If you encounter issues with pnpm v7+ (not running postinstall scripts), follow the manual registration steps outlined in the README to ensure the bridge is registered correctly.
• The toolset includes browser management commands (get_windows_and_tabs, chrome_navigate, chrome_switch_tab, chrome_close_tabs, chrome_go_back_or_forward, chrome_inject_script, chrome_send_command_to_inject_script) and a rich collection of other browser-related tools referenced in the Complete Tool List.