home / mcp / browser control mcp server
MCP server paired with a browser extension that enables AI agents to control the user's browser.
Configuration
View docs{
"mcpServers": {
"eyalzh-browser-control-mcp": {
"command": "node",
"args": [
"/path/to/repo/mcp-server/dist/server.js"
],
"env": {
"EXTENSION_PORT": "8089",
"EXTENSION_SECRET": "<secret_on_firefox_extension_options_page>"
}
}
}
}Browser Control MCP is a server that runs locally alongside a Firefox extension to give AI assistants safe access to your browser’s tabs, history, and page content. It lets you manage tabs, search history, and read page content with your explicit consent, helping you research and navigate more efficiently while keeping control over what the assistant can do.
You connect an MCP client to the Browser Control MCP to perform tasks like opening or closing tabs, grouping and reordering tabs, reading your history, and reading or highlighting text on web pages. Each action is initiated from your MCP client, and the extension asks for consent when reading page content. You can also review an in-extension audit log to see which tool calls were made and how they were configured.
Prerequisites: Node.js and npm if you build and run the MCP server from source. A running Firefox browser with the Browser Control MCP extension installed is also required.
Option 1: Install the Firefox extension and Claude Desktop integration (recommended if you want a ready-to-use setup). Follow the extension installation steps, then install and enable the DXT extension in Claude Desktop. Configure the MCP server as described below.
Option 2: Build from code and run the MCP server locally. Use the following steps to build and start the server and extension.
# Install dependencies and build both MCP server and browser extension
npm install
npm run buildAfter you install the browser extension, configure the MCP server to connect to the extension. Use the following configuration snippet as a starting point. Fill in your actual secret value from the extension preferences.
{
"mcpServers": {
"browser-control": {
"command": "node",
"args": [
"/path/to/repo/mcp-server/dist/server.js"
],
"env": {
"EXTENSION_SECRET": "<secret_on_firefox_extension_preferences>",
"EXTENSION_PORT": "8089"
}
}
}
}If you prefer running the MCP server in Docker, build the image and run it with the required environment variables.
# Build the Docker image
docker build -t browser-control-mcp .
# Run the container with the required env vars and port mapping
docker run --rm -i -p 127.0.0.1:8089:8089 -e EXTENSION_SECRET=<secret_from_extension> -e CONTAINERIZED=true browser-control-mcpThis MCP server operates locally and communicates with the extension over a local connection using a shared secret. It does not modify web pages by itself and requires explicit user consent for reading page content. Use the in-extension audit log to review tool calls and enable/disable configurations as needed. Exercise caution with tool calls and monitor new or unfamiliar uses.
The Browser Control MCP exposes a set of browser-related actions that your AI assistant can request. These include managing tabs, querying history, and reading page content (with user consent). You can also search and highlight text within a tab. Each tool is described in detail in the Tools section of your client’s integration guide.
Open or close browser tabs to manage your workflow.
Retrieve the list of currently opened tabs in the browser.
Create named tab groups and assign colors for visual organization.
Change the order of open tabs to suit your preferences.
Read and search through the browser history.
Read a webpage's text content and links (requires user consent per domain).
Find and highlight specific text within a browser tab (requires user consent).