PackForAI MCP server

Convert PDF, DOCX, PPTX, XLSX, CSV and JSON into clean, compact, AI-ready Markdown, directly inside Claude, Cursor and any other MCP client. Powered by PackForAI.

Instead of pasting a messy PDF into your AI (broken text, lost tables, wasted tokens), ask your assistant to convert it first: it comes back as clean Markdown with up to 65% fewer tokens.

What you need

• Node.js 18+
• A PackForAI API key (Pro plan). Create one at packforai.com/account.

Install

The server runs via npx, so there is nothing to install globally. Add it to your MCP client's config.

Claude Desktop

Edit claude_desktop_config.json (Settings → Developer → Edit Config) and add:

{
  "mcpServers": {
    "packforai": {
      "command": "npx",
      "args": ["-y", "packforai-mcp"],
      "env": { "PACKFORAI_API_KEY": "your_key_here" }
    }
  }
}

Restart Claude Desktop.

Cursor

Add to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "packforai": {
      "command": "npx",
      "args": ["-y", "packforai-mcp"],
      "env": { "PACKFORAI_API_KEY": "your_key_here" }
    }
  }
}

Usage

Once connected, just ask your assistant, for example:

• "Convert ~/Downloads/Q4-report.pdf to clean Markdown."
• "Fetch https://example.com/spec.docx and give me the AI-ready version."
• "Convert this scanned PDF with OCR: /path/to/scan.pdf."

Tool

convert_document

Param	Type	Description
path	string	Absolute path to a local document file.
url	string	Public https URL of a document to fetch and convert.
ocr	boolean	Force OCR for scanned / image-only PDFs (Pro). Default false.
full	boolean	Return the full document.md (e.g. complete spreadsheet tables) instead of the compact version. Default false.

Provide either path or url. Returns the Markdown plus a token-savings summary.

Config

• PACKFORAI_API_KEY (required) — your PackForAI API key.
• PACKFORAI_BASE_URL (optional) — override the API base (defaults to https://packforai.com/api/v1).

Links

• Website: packforai.com
• API docs: packforai.com/docs

License

MIT