home / mcp / browser manager mcp server

Browser Manager MCP Server

Browser Manager MCP Server est un serveur MCP (Model Context Protocol) spécialisé dans la gestion et l automatisation des navigateurs web. Il fournit une interface complète pour contrôler les navigateurs, gérer les onglets, interagir avec les pages web et effectuer des tâches d'automatisation.

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "deamondev888-browser-manager-mcp-server": {
      "url": "http://localhost:8081/mcp",
      "headers": {
        "BEARER_TOKEN": "YOUR_TOKEN"
      }
    }
  }
}

Browser Manager MCP Server lets you control and automate web browsers via the MCP (Model Context Protocol). It provides a unified interface to launch and manage browsers, open and switch tabs, automate interactions, capture data, and stream real-time communications. This makes it easy to build automated workflows that interact with any Chromium-based browser from your MCP client.

How to use

You connect to the MCP server from an MCP client and issue high-level actions to control browsers and their pages. You can launch browsers, list and manage tabs, navigate to URLs, perform clicks and text input, and capture page data or screenshots. Real-time updates are available via streaming endpoints, and optional Bearer token authentication protects access.

How to install

# Prerequisites
- Node.js 24.0.2 or newer
- npm 11.3.0 or newer (recommended)

# 0. Download the project
# clone the complete repository
git clone https://github.com/Jboner-Corvus/Browser-Manager-MCP-Server.git
# Enter the dist folder
cd "Browser-Manager-MCP-Server/dist"

# 1. Install dependencies
npm install --production

# 2. Configure the environment
# The .env file is present with default values. Edit as needed.
```

Then you can start the server with one of the following methods.
# Recommended startup on Windows
.\start-prod.ps1

# Manual startup
node server.js
```

If you want to run using the stdio transport (for MCP clients that use standard input/output), set MCP_TRANSPORT accordingly and start the server with the final command shown above. For example, to run in stdio mode you would configure the start command as shown in the example below.

Additional sections

Endpoints and transport options expose how you connect to and stream data from the server.

- Primary MCP endpoint: http://localhost:8081/mcp (HTTP Stream, JSON-RPC) with optional Bearer authentication. - Server-Sent Events endpoint: http://localhost:8081/sse (Streaming) with optional Bearer authentication. - Transport modes: - HTTP Stream (default): node lib/server.js, endpoints /mcp and /sse. - Stdio: MCP_TRANSPORT=stdio node lib/server.js (command uses node and lib/server.js). - SSE: MCP_TRANSPORT=sse node lib/server.js (redirects /mcp to streaming). If you need to connect with external browsers via an extension, install and use the included Browser Manager MCP Bridge extension. This extension communicates with the server over a WebSocket relay on port 8082 and supports Chromium-based browsers.

Notes on capabilities and usage

The server provides comprehensive browser management, tab handling, navigation, interaction automation, and data capture. Use the available tools to orchestrate complex automation flows across one or multiple browsers and tabs. Authentication is optional; enable it to restrict access using a Bearer token.

Available tools

launch_browser

Launch a new browser instance to be managed by the MCP server.

list_browsers

List all browsers currently managed by the server, including their active tabs.

close_browser

Close a specified browser instance.

detect_open_browsers

Detect browsers currently running on the system.

list_tabs

List all open tabs within a browser.

select_tab

Select a specific tab to be the active focus for subsequent actions.

new_tab

Open a new tab in the selected browser.

close_tab

Close a specified tab.

navigate

Navigate the active tab to a given URL.

click

Perform a click action on a page element.

type_text

Type text into an input field on the page.

wait_for

Wait for specific text to appear or for a timeout.

get_html

Retrieve the HTML content of the current page.

get_console_logs

Fetch console logs from the browser for debugging.

screenshot

Capture a screenshot of the current viewport.

evaluate_script

Execute JavaScript in the context of the current page.