Whistler MCP Server

An official Model Context Protocol (MCP) server for interacting with the Whistler placement data platform. This allows AI assistants like Claude Desktop, Cursor, and others to seamlessly read job roles, platform statistics, and PDFs on your behalf.

Features

• Job Placement Data: Fetch lists of companies and specific job roles for any requested year.
• PDF Reading: Natively extracts text from job description PDFs (respects your daily view limit).
• Feedback & Requests: Submit bug reports, feature ideas, and PDF limit increase requests dynamically.
• Agentic Authentication: The AI securely prompts you for your email and password when you first connect, fetches a real token from the backend, and saves it locally (~/.whistler_mcp_token.json).

Usage via npx (Claude Desktop)

To use this server locally with Claude Desktop, add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "whistler-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "whistler-mcp"
      ],
      "env": {
        "API_URL": "https://whistler-backend.onrender.com/api"
      }
    }
  }
}

Note: You do not need to provide a JWT token in the config. Try asking Claude a Whistler-related question, and it will prompt you to securely log in directly within the chat interface!

Local Testing & Debugging

If you want to test the server without an AI client, you can use the official MCP Inspector:

npx -y @modelcontextprotocol/inspector npx -y whistler-mcp

This will open a local web debugger at http://localhost:5173 where you can manually run and inspect each tool.

Distribution

This folder is a completely standalone Node.js package. It has its own internal dependencies and maps to the Whistler API over HTTP. You can safely move this folder anywhere on your computer (or even to a brand new repository) and it will continue to function perfectly as long as your API_URL is accessible.

Local Development

1. Clone the repository.
2. Run npm install inside the /whistler-mcp folder.
3. Test locally using the node server.js command.