pdf-to-data-mcp

An MCP server that lets AI agents (Claude Desktop, Claude Code, Cursor, and any MCP client) convert PDFs and images into structured spreadsheet data — invoices, bank & brokerage statements, utility bills, purchase orders, and more.

It exposes one tool, convert_document, which sends a local file to the converter API and returns clean JSON rows.

Requirements

• Node.js 18+
• An account with an active Pro subscription (or credits) on one of the converters, and your API key from Settings → Your API key:
  • freightinvoicetoexcel.com
  • utilitybillextractor.com
  • purchaseorderconverter.com
  • brokeragestatementconverter.com

Configuration

Add it to your MCP client. For Claude Desktop, edit claude_desktop_config.json:

{
  "mcpServers": {
    "brokerage-converter": {
      "command": "npx",
      "args": ["-y", "pdf-to-data-mcp"],
      "env": {
        "CONVERTER_BASE_URL": "https://brokeragestatementconverter.com",
        "CONVERTER_API_KEY": "api-YOUR_KEY_HERE"
      }
    }
  }
}

CONVERTER_BASE_URL selects which converter to use — swap it for any of the four domains. To use several at once, add multiple entries (e.g. freight-converter, utility-converter) each with its own base URL and key.

Usage

Once configured, just ask your agent naturally:

"Convert ~/Downloads/statement.pdf and give me the transactions as a table."

The agent calls convert_document with the file path and gets back the extracted columns and rows.

Tool

convert_document

• file_path (string, required) — absolute path to a local PDF, PNG, JPG, or WebP file (max 20 MB).
• Returns JSON: { filename, rowCount, columns, rows }.

Publishing (maintainers)

cd mcp-server
npm install
npm publish --access public

Then list it in the MCP registries for discovery: mcp.so, Smithery, and the official modelcontextprotocol/registry.

License

MIT