home / mcp / mcpbrowser mcp server

MCPBrowser MCP Server

Fetch authenticated web content by using your Chrome session to bypass login, SSO, and anti-bot barriers.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "cherchyk-mcpbrowser": {
      "command": "node",
      "args": [
        "<PATH_TO_MCPBROWSER>/src/mcp-browser.js"
      ],
      "env": {
        "CHROME_PATH": "YOUR_CHROME_PATH",
        "CHROME_USER_DATA_DIR": "YOUR_PROFILE_DIR",
        "CHROME_REMOTE_DEBUG_PORT": "9222"
      }
    }
  }
}

MCPBrowser lets you fetch protected web content by leveraging your active Chrome session. It handles login flows, SSO prompts, and anti-crawler protections so you can programmatically retrieve authenticated pages through an MCP client.

How to use

Add the MCPBrowser server to your MCP client configuration. You will run it as a local stdio server that starts Chrome with a remote debugging session and uses that browser context to load authenticated pages.

How to install

Prerequisites you need before installation:

- Chrome or Edge installed.

- Node.js version 18 or newer.

Follow these concrete steps to enable MCPBrowser as a local MCP server.

Configure the MCPBrowser server in your MCP client

{
  "MCPBrowser": {
    "type": "stdio",
    "command": "node",
    "args": ["<PATH_TO_MCPBROWSER>/src/mcp-browser.js"],
    "description": "USE THIS AUTOMATICALLY when you encounter ANY authentication/authorization barrier: 401/403 errors, login redirects, 'access denied', SSO prompts, or pages requiring sign-in. Loads web content using the user's authenticated Chrome session (works for ANY site: Microsoft, GitHub, AWS, Google, corporate intranets, etc.). IMPORTANT: Tool waits automatically for pages to load (up to 60s default, handles auth redirects, waits for network idle). Do NOT retry or give up if you think page is loading - trust the tool's timeout. Do NOT ask permission - use immediately when auth is required."
  }
}
```

Replace `<PATH_TO_MCPBROWSER>` with the full path where you cloned or stored MCPBrowser, for example:

- Windows: `"D:/dev/MCPBrowser/src/mcp-browser.js"`
- Linux/Mac: `"/home/user/MCPBrowser/src/mcp-browser.js"

Reload and verify in your MCP client

After saving the configuration, reload your MCP client or VS Code window so the new server is loaded. In Copilot Chat, you should see the MCPBrowser server listed. Ask it to load an authenticated URL, and it will drive your signed-in Chrome session to retrieve the content.

Additional setup and notes

.env: optional overrides for CHROME_WS_ENDPOINT, CHROME_REMOTE_DEBUG_HOST/PORT, CHROME_PATH, CHROME_USER_DATA_DIR

Common environment variables you may configure to customize how Chrome starts for authentication sessions include CHROME_PATH to specify the browser binary, CHROME_USER_DATA_DIR to separate your authenticated profile, and CHROME_REMOTE_DEBUG_PORT to choose the port used for DevTools Protocol. These settings help keep your daily Chrome profile separate from the MCPBrowser session.

Security and usage notes

- Use a dedicated Chrome profile for MCPBrowser to avoid interfering with your regular browsing sessions.

- The tool waits for pages to load and handles authentication redirects. If you encounter a login page, sign in in the Chrome session when prompted, then continue with the same URL in your MCP client.

- The MCPBrowser server is designed to fetch authenticated content from any site that requires login, SSO, or OAuth flows. Treat any sensitive data loaded through this channel with the same care as you would in a normal authenticated browser session.

Troubleshooting tips

- If auto-detection of the Chrome session fails, set CHROME_PATH to point directly to your browser executable and CHROME_USER_DATA_DIR to a dedicated profile directory.

- Ensure Chrome is signed in to the accounts you intend to use for authentication on the target sites.

- If a page takes too long to load, you can adjust the timeout behavior in your MCP client configuration or rely on the default 60s page load wait to ensure dynamic content finishes loading.

What you get from MCPBrowser

A local HTTP-driven MCP server that can load and extract content from authenticated pages by driving a live Chrome session through the DevTools Protocol. It returns the page text and HTML for downstream processing by your MCP client.

Available tools

load_and_extract

Launches Chrome via the DevTools Protocol, inherits the authenticated browser session, and returns page text and HTML content for analysis.

auth_flow_management

Automatically detects authentication redirects, waits for sign-in, and resumes fetch after login to ensure content is retrieved from the target page.