mcp-web-browser

This MCP server provides a headless web browsing API powered by Playwright. It enables you to programmatically navigate websites, extract full HTML content (including dynamic JavaScript), manage multiple tabs, interact with page elements (text input, clicks), capture screenshots, and retrieve page metadata. The API supports operations such as browse_to, extract_text_content, input_text, click_element, get_page_screenshots, get_page_links, create_new_tab, switch_tab, list_tabs, close_tab, refresh_page, wait_for_navigation, execute_javascript, and get_page_info. It’s designed for automated workflows, data extraction, and browser-based testing within the MCP framework. Use multiple tabs to parallelize tasks or isolate sessions, and leverage the built-in cleanup to release resources after inactivity.

Prerequisites:

• Python 3.10 or newer
• Network access to install Python packages

1. Create and activate a Python virtual environment (recommended):

python -m venv venv
# macOS/Linux
source venv/bin/activate
# Windows
venv\Scripts\activate

2. Install the MCP framework and Playwright bindings:

pip install mcp playwright

3. Install browser dependencies for Playwright:

playwright install

4. Run the MCP Web Browser server (example):

python /path/to/your/server.py

5. Optional: Configure your Claude Desktop or other MCP client to point to the server using the provided MCP config format (see mcp_config example).

Tips and common considerations:

• Ensure Python 3.10+ is installed and available in your PATH.
• The server bypasses SSL certificate validation for web browsing; use with awareness of security implications.
• The Playwright browser context can be configured for secure operation; adjust user-agent and timeouts as needed via the API.
• If you encounter slow page loads, tune timeouts in browse_to and related navigation methods.
• Resources are automatically cleaned up after inactivity; if you expect long-running sessions, monitor resource usage.
• When running behind a firewall, ensure that required ports for MCP communication are accessible.
• For troubleshooting, enable verbose logging in your MCP client and review server logs for navigation errors or selector issues.