auth-fetch-mcp

MCP server that lets AI assistants fetch content from authenticated web pages.

When your AI tries to read a URL that requires login, this tool opens a real browser for you to sign in — then captures the page content. Sessions are saved locally, so you only log in once per service.

Demo

"Summarize this Notion page for me"

A browser opens with a Capture button. Log in if needed, then click it:

The AI receives the full page content and responds:

Quick Start

Claude Code

claude mcp add --scope user auth-fetch -- npx auth-fetch-mcp@latest

.mcp.json (Cursor, Windsurf, etc.)

{
  "mcpServers": {
    "auth-fetch": {
      "command": "npx",
      "args": ["auth-fetch-mcp@latest"]
    }
  }
}

Chromium is auto-installed on first run if not already present.

How It Works

1. Ask your AI to read any authenticated page — just paste the URL.

2. A browser window opens automatically and navigates to the page.

3. Log in as you normally would (supports SSO, 2FA, CAPTCHA — anything).

4. Click the "📸 Capture" button in the bottom-right corner when ready.

5. The page content is captured, the browser closes, and your AI receives the content.

Tools

auth_fetch

The primary tool. Fetches page content using a real browser, opening a window for login if needed.

Flow:

1. Opens a headed browser and navigates to the URL

2. A floating "📸 Capture" button appears on the page

3. User logs in or navigates as needed (button re-appears after page transitions)

4. User clicks the button — content is captured as Markdown — browser closes

list_pages

Lists all open tabs in the browser with their URLs and titles.

close_browser

Closes the browser window. Login sessions are saved and will be reused next time.

How Sessions Work

Login sessions are saved to ~/.auth-fetch-mcp/browser-data/. This is a standard Chromium profile directory containing cookies and local storage. Sessions persist across restarts, so previously logged-in sites will load faster on the next visit.

To clear all sessions:

rm -rf ~/.auth-fetch-mcp/browser-data/

Supported AI Tools

• Claude Code

• Cursor

• Windsurf

• Any MCP-compatible client using stdio transport

Limitations

• Requires a local environment (does not work in web-based chat interfaces)

• First access to each service requires manual login

• Very long pages are truncated to fit LLM context windows (50K chars)

• Some sites with aggressive bot detection may not work (try the wait_for option)

Privacy

• All data stays on your machine — nothing is sent to external servers

• Browser sessions are stored locally in your home directory

• The MCP server only communicates with the AI tool via stdio (local pipe)

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

git clone https://github.com/ymw0407/auth-fetch-mcp.git
cd auth-fetch-mcp
npm install
npm run build

License

MIT