mcpX product documentation
Get started with mcpX
Connect an MCP-compatible AI client to Chrome and run your first browser task.
mcpX gives an MCP-compatible AI client a controlled set of Chrome tools. It can inspect a page, navigate, click, type, wait for changes, capture screenshots, and collect console output through a server that runs on your computer.
What you need
- Chrome 120 or newer
- Node.js 20 or newer
- An MCP-compatible client such as Cursor, Claude Desktop, VS Code, or Windsurf
- The mcpX Chrome extension
No mcpX account or hosted relay is required.
Set up in three steps
1. Add the MCP server
Add this server entry to your AI client:
{
"mcpServers": {
"browser-mcpx": {
"command": "npx",
"args": ["-y", "@rpx_/mcpx@latest"]
}
}
}Restart or reload the client after saving its MCP configuration.
2. Connect Chrome
Install the extension, open it from the Chrome toolbar, and select Connect browser. The extension discovers the local server and pairs automatically.
3. Run a browser task
Open a normal web page and ask your client:
Use mcpX to list my open tabs, inspect the active page, and summarize its main sections.Your client should call tabs-list or tab-current, then snapshot.
Where to go next
| Goal | Guide |
|---|---|
| Configure the npm server | Set up the MCP server |
| Install or reconnect Chrome | Connect the extension |
| Learn reliable prompting patterns | Run your first workflow |
| See every available browser action | Tool reference |
| Understand what can leave a page | Security and privacy |
If the extension does not connect, start with Troubleshooting.