Chrome MCP Server
A Chrome extension with an MCP server: the agent drives your everyday browser with its sessions, tabs, bookmarks and history
Install
npm install -g mcp-chrome-bridgeFirst download the extension from releases and load it unpacked at chrome://extensions.
This is third-party code. Review the repository files before installing.
What it does
The project is a Chrome extension plus the mcp-chrome-bridge that exposes browser capabilities to an MCP client. Unlike Playwright-based tools, the agent works in your already open browser where you are logged in and configured. It offers navigation and tab switching, screenshots, network capture, page content extraction, clicks, form filling, bookmarks and history. A built-in vector store enables semantic search across open tabs.
Who it is for. For people who need an agent to work on websites under their own account without logging in again.
Good fit when
- The agent needs sites where you are already signed in
- You need to gather data across several open tabs
- You want to inspect a page's network requests and console
Not a fit when
- You need isolated automation for tests or CI, where Playwright fits better
- The browser holds banking or work sessions the agent must not touch
Example request
Open my MCP bookmarks, go through the pages and write a short summaryLimitations
The project is early stage and the authors warn about instability. The extension is loaded manually in developer mode from GitHub releases. Requires Node.js 20 or newer. The agent gets access to all sessions and data in your browser.
How to disable. Remove the chrome-mcp-server entry from the client's MCP config, disable or remove the extension at chrome://extensions and run npm uninstall -g mcp-chrome-bridge.
MCP
- Transport
- http, stdio
- Authentication
- not required
Security check
- The agent acts in the browser under your accounts
- Can read history, bookmarks and network traffic and inject scripts into pages
README in short
The README compares the extension approach with Playwright-based servers and lists features. Setup has three steps: download the extension from releases, install mcp-chrome-bridge globally and load the extension in Chrome. It gives configs for Streamable HTTP and stdio and a full grouped tool list. MIT licensed.
FAQ
How is this different from Playwright MCP?
No separate browser is launched: the agent uses your Chrome with its settings and signed-in accounts.
What if my client only supports stdio?
The README shows how to find mcp-server-stdio.js in global packages and point the config at it.
Related
CLI, skill and Python library for browser control: the agent clicks, fills forms and reads pages over CDP
A skill that gathers the last 30 days of discussion on a topic from Reddit, X, YouTube, HN, Polymarket and GitHub into one brief
The Chrome DevTools team's official MCP server: the agent drives a live Chrome, reads network and console, and records performance traces
A fast Rust CLI for agent browser automation: accessibility snapshots with element refs, an MCP server and skills