Guide · Open source

How to use mcp-chrome: install the bridge and connect Chrome to an AI assistant

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

mcp-chrome (Chrome MCP Server) is an open-source Chrome extension paired with a local Model Context Protocol server. It lets an AI assistant such as Claude work inside the Chrome browser you already use, with your existing logins and settings. This guide shows you how to install the bridge, load the extension, connect an MCP client and understand what the tools can do.

Chrome MCP Serverhangwin/mcp-chrome · ★ 12.5k · MIT Install options & downloads

Quick start

1Check prerequisites. Confirm Node.js 20.0.0 or newer, pnpm or npm, and Chrome or Chromium are installed.
2Download the extension. Get the latest Chrome extension from the project's GitHub releases page.
3Install the bridge. Run npm install -g mcp-chrome-bridge, or the pnpm equivalent after enabling pre/post scripts.
4Register if needed. If postinstall did not run, run mcp-chrome-bridge register.
5Load the extension. In chrome://extensions/, turn on Developer mode, then Load unpacked and choose the extracted folder.
6Connect your MCP client. Add the Streamable HTTP entry with url http://127.0.0.1:12306/mcp, or use the STDIO config if your client requires it.

What you need before you start

The project's README lists the prerequisites: Node.js version 20.0.0 or newer, pnpm or npm, and a Chrome or Chromium browser.

Node.js >= 20.0.0
pnpm or npm
Chrome or Chromium browser
An MCP-compatible client, such as Claude or CherryStudio (the README uses CherryStudio as its example)

Install the mcp-chrome-bridge package

First, download the latest Chrome extension from the project's releases page on GitHub. Then install the bridge globally with npm or pnpm. With pnpm, the README recommends enabling pre/post install scripts first, because pnpm v7+ disables postinstall scripts by default.

If postinstall did not run, register the bridge manually with the register command shown below.

npm install -g mcp-chrome-bridge

# pnpm (recommended method)
pnpm config set enable-pre-post-scripts true
pnpm install -g mcp-chrome-bridge

# pnpm manual registration, if postinstall doesn't run
pnpm install -g mcp-chrome-bridge
mcp-chrome-bridge register

Load the Chrome extension

Open Chrome and go to chrome://extensions/. Turn on Developer mode, click Load unpacked, and select the folder where you extracted the downloaded extension. Then click the extension icon and click connect to see the MCP configuration.

Open chrome://extensions/
Enable Developer mode
Click Load unpacked and select your extension folder
Click the extension icon, then connect

Connect an MCP client using Streamable HTTP

The README recommends Streamable HTTP. Add the following entry to your MCP client configuration. The server listens at the URL shown in the config.

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

Use the STDIO connection if your client needs it

If your client only supports stdio, find where the npm package was installed using the list command for your package manager. Then point the config at the mcp-server-stdio.js file inside that install path. Replace the example path with the one your own machine reports.

Run npm list -g mcp-chrome-bridge or pnpm list -g mcp-chrome-bridge to find the install path
Append node_modules/mcp-chrome-bridge/dist/mcp/mcp-server-stdio.js to that path
Put the full path into the args array of your client config

npm list -g mcp-chrome-bridge
pnpm list -g mcp-chrome-bridge

{
  "mcpServers": {
    "chrome-mcp-stdio": {
      "command": "npx",
      "args": [
        "node",
        "/Users/xxx/Library/pnpm/global/5/node_modules/mcp-chrome-bridge/dist/mcp/mcp-server-stdio.js"
      ]
    }
  }
}

What the tools can do

The server exposes more than 20 tools, grouped by purpose. Examples include listing windows and tabs, navigating, taking screenshots, capturing network traffic, extracting page content, clicking elements by CSS selector and filling forms. The full list is in docs/TOOLS.md in the repository.

Browser management: get_windows_and_tabs, chrome_navigate, chrome_switch_tab
Screenshots: chrome_screenshot, with element targeting and full-page support
Network monitoring: chrome_network_capture_start/stop, chrome_network_debugger_start/stop
Content analysis: search_tabs_content (semantic search across tabs), chrome_get_web_content, chrome_console
Interaction: chrome_click_element, chrome_fill_or_select

What people use it for

Inspecting or testing a page in your everyday Chrome profile with an AI assistant
Letting an assistant work inside sites where you are already logged in
Capturing screenshots and console output from open tabs
Running semantic search across the content of your open browser tabs
Using browser tools from an MCP-compatible client without launching a separate browser

Frequently asked questions

Is mcp-chrome free and open source?

Yes. The project is licensed under the MIT License, and its source is written in TypeScript. The README does not mention any paid tier.

What do I need to run mcp-chrome?

You need Node.js 20.0.0 or newer, pnpm or npm, and a Chrome or Chromium browser. You also need an MCP-compatible client to connect to the server.

Why won't the bridge register after installing with pnpm?

pnpm v7+ disables postinstall scripts by default. Enable pre/post install scripts with pnpm config set enable-pre-post-scripts true, or run mcp-chrome-bridge register manually after installing.

How is mcp-chrome different from Playwright MCP?

Playwright-based servers launch a separate browser with a clean environment, so you must log in again. mcp-chrome uses your existing Chrome browser and login states, and the README says it avoids launching an independent browser process.

Is it safe to let an AI control my browser?

The README notes that the server runs locally for privacy. However, because the assistant acts inside a browser with active login sessions, you should only connect clients you trust and be careful about which sites are open.

Is mcp-chrome ready for production use?

No. The README says the project is in early stages and under intensive development, so it is not a stable, production-grade tool.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

mcp-chrome suits developers and everyday Chrome users who want an AI assistant to inspect pages or work inside sites they are already logged into. Skip it if you need isolated, repeatable test runs or a stable production tool, since the project says it is still in early stages. For clean test environments, a Playwright-based MCP server is the better fit.

Install options & downloads