home / mcp / browser use mcp server

Browser Use MCP Server

Browse the web, directly from Cursor etc.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "kontext-dev-browser-use-mcp-server": {
      "url": "http://localhost:8000/sse",
      "headers": {
        "CHROME_PATH": "/usr/bin/google-chrome",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

You run a browser-control MCP server that lets AI agents operate web browsers through automated actions. It supports two transport modes (SSE for direct server events and stdio for local processes) and can stream real-time browser activity so you can observe tasks as they happen. This empowers you to build agents that navigate sites, extract information, and perform actions across web pages with confidence.

How to use

Choose your preferred transport and connect your MCP client to the server. In SSE mode, you run the server and point your client at the SSE endpoint to receive events and send commands. In stdio mode, you launch the local server process and use a stdio transport to communicate with it from the client.

How to install

Prerequisites are required before you start. You need a Python-compatible runtime managed by the UV tool, and you should have Playwright installed for browser automation. If you intend to use stdio mode, you will also rely on mcp-proxy.

# Install prerequisites
curl -LsSf https://astral.sh/uv/install.sh | sh
uv tool install mcp-proxy
uv tool update-shell

# Install Playwright and browsers
uv sync
uv pip install playwright
uv run playwright install --with-deps --no-shell chromium

# Optional: configure environment for API access
# Create a .env file in your project root
OPENAI_API_KEY=your-api-key
CHROME_PATH=optional/path/to/chrome
PATIENT=false  # Set to true if API calls should wait for task completion

Usage details

SSE mode lets you run the server directly. Start the server on a port of your choice and connect your MCP client to the provided SSE URL.

stdio mode runs a local process that communicates over a stdio channel. You build and install the server as a Python wheel, then launch it with a command that enables stdio transport and exposes a proxy port for the client to connect.

Environment

Create a .env file with your configuration. At minimum you provide your OpenAI API key and can optionally specify where Chrome is located and whether tasks should be synchronous.

OPENAI_API_KEY=your-api-key
CHROME_PATH=optional/path/to/chrome
PATIENT=false  # Set to true if API calls should wait for task completion

Examples of common flows

- Start in SSE mode and connect your client to http://localhost:8000/sse to perform and observe tasks.

- Build the local stdio server and run it so your client can communicate over a stdio transport with a proxy port.

Available tools

browser_automation

Control browsers programmatically, enabling AI agents to perform navigation, form filling, clicks, and data extraction.

vnc_streaming

Watch browser automation in real-time via a VNC-like viewer for live debugging and verification.

async_tasks

Execute browser operations asynchronously, allowing tasks to run in parallel and notify when complete.