MCP2Browser

You can automate and inspect web pages through a dedicated MCP server over HTTP. This server lets you open web pages and retrieve page titles using Playwright, enabling AI assistants to interact with browsers programmatically in a lightweight, HTTP-based MCP workflow.

How to use

You connect to the browser MCP server from an MCP client and issue actions such as opening a page or reading its title. Start by ensuring the server is running, then configure your MCP client to communicate with the server URL shown below. Use the available actions to drive browser interactions in your automation or AI-assisted workflows. For common tasks, open a page to navigate to a URL and then request the title of the current page. You can combine actions into a single workflow, such as opening a page and then retrieving its title.

How to install

Prerequisites: Node.js v16 or higher and npm or yarn.

Step 1: Clone the project and enter the directory.

git clone <repository-url>

cd mcp2browser

Step 2: Install dependencies.

npm install

Step 3: Build the project.

npm run build

Step 4: Start the server.

npm start

The server will start running on http://localhost:3000/mcp.

Configuration and usage notes

Configure your MCP client to connect to the HTTP MCP server at the designated URL. The server uses an Express-based MCP implementation with Playwright for browser automation.

Example: Ensure the server is reachable at the following address before starting client sessions: http://localhost:3000/mcp

VS Code integration and examples

If you are using the integrated Copilot chat in VS Code, configure the MCP server so Copilot can send browser commands. You can add the server either through the command palette or by editing your settings.

Option 1: Using MCP: Add Server in VS Code

Option 2: Manual configuration in JSON

For a manual setup, add an MCP server with the following configuration snippet in your settings.json, replacing the URL if needed.

Example configuration for VS Code

{
  "github.copilot.chat.mcp.servers": {
    "browser_mcp": {
      "type": "http",
      "url": "http://localhost:3000/mcp",
      "transport": "streamable-http"
    }
  }
}

Notes on usage and capabilities

Available actions include opening a web page and retrieving the page title. Use these actions to build practical browser automation workflows with your MCP client. Ensure the server is running before attempting to connect from your client, and verify the correct server URL is configured in your MCP settings.

Troubleshooting

Issue: Server not connecting in the MCP client. Make sure the server is running at http://localhost:3000/mcp and that the client is configured to use that URL. Restart the client after changing settings.

Issue: Tools not appearing in the MCP client. Verify the server URL and transport settings in your client configuration, and check for any server-side startup errors in the console.

Available tools

openPage

Opens a web page in a headless browser by navigating to a specified URL.

getPageTitle

Retrieves the title of the currently opened web page.