PDFMakerAPI MCP server

An MCP server for PDFMakerAPI. It exposes a single create_document tool that turns a plain-English document description into a shareable link which opens in the PDFMakerAPI editor — preview it, edit any field, and download the PDF.

Works with Claude Desktop, Claude.ai, Cursor, Windsurf, Cline, Zed, VS Code, ChatGPT, and any other MCP client. It's a thin client of the public API at https://api.pdfmakerapi.com, so it needs no account or credentials of its own.

What is PDFMakerAPI?

PDFMakerAPI is a PDF template builder and PDF generation API. Instead of hard-coding documents or wrestling with HTML-to-PDF converters that break when your content changes, you design a reusable template once in a visual, no-code editor — placing text, tables, images, and dynamic fields exactly where you want them — then feed it data to produce a finished, editable PDF.

Generate a single document or thousands. Fill a template in yourself, share a link so anyone completes it in the browser as a fillable PDF form, send JSON to the REST API, trigger generation from no-code tools like Zapier, Make, and n8n, or create documents straight from AI agents like Claude and ChatGPT. Prefer to start from a prompt? Describe what you need and let Quill build it.

Teams use PDFMakerAPI for invoices, receipts, certificates, reports, and contracts — anything you'd otherwise rebuild by hand every time the data changes. Start building free with 100 PDFs a month.

Quick install

One-click:

Or add the same config manually (works in every desktop client):

{
  "mcpServers": {
    "pdfmakerapi": {
      "command": "npx",
      "args": ["-y", "@pdfmakerapi/mcp"]
    }
  }
}

Setup by client

Settings → Developer → Edit Config (claude_desktop_config.json), add the mcpServers block above, and restart Claude.

Use the Add to Cursor button, or add the block to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project).

Add the block to ~/.codeium/windsurf/mcp_config.json (or via Settings → Cascade → MCP).

Cline → MCP Servers → Configure, and add the mcpServers block.

Use the Install in VS Code button, or add to .vscode/mcp.json (note: VS Code uses a servers key):

{
  "servers": {
    "pdfmakerapi": { "command": "npx", "args": ["-y", "@pdfmakerapi/mcp"] }
  }
}

In settings.json:

{
  "context_servers": {
    "pdfmakerapi": { "command": { "path": "npx", "args": ["-y", "@pdfmakerapi/mcp"] } }
  }
}

Hosted (remote) option — no install

If your client supports remote MCP servers, just point it at the hosted endpoint — no npx, no Node:

https://api.pdfmakerapi.com/mcp

• Claude.ai (web): Settings → Connectors → Add custom connector → paste the URL.
• ChatGPT (Plus/Pro/Enterprise): Settings → Connectors → add the URL.
• Cursor / others: add { "url": "https://api.pdfmakerapi.com/mcp" } instead of command/args.

Usage

Ask your assistant for a document:

"Make a professional invoice for Acme with 3 line items." "Create a course completion certificate." "Build a clean one-page resume."

It calls create_document and returns a link like https://app.pdfmakerapi.com/d/<id> — open it to preview, edit any field, and download the PDF.

Configuration

| Env var | Default | Purpose | | --- | --- | --- | | PDFMAKERAPI_BASE_URL | https://api.pdfmakerapi.com | Point at a different API (e.g. http://localhost:3001 for local dev). |

Develop

git clone https://github.com/GerardoBarrera/pdfmakerapi-mcp.git
cd pdfmakerapi-mcp
npm install
npm run build   # compile to dist/
npm run dev     # run from source (tsx)

Troubleshooting

• Server doesn't appear in your client — restart the client after editing its MCP config, and check the JSON is valid (no trailing commas). For remote clients, confirm the URL is exactly https://api.pdfmakerapi.com/mcp.
• npx fails to start — ensure Node.js 18+ is installed (node -v), then retry npx -y @pdfmakerapi/mcp@latest.
• Connection / timeout on the remote endpoint — verify your network/proxy allows HTTPS to api.pdfmakerapi.com, and that you're using the Streamable HTTP transport.
• "Document is too large" — documents are capped at ~1 MB of JSON; trim large tables or split into multiple documents.
• The returned link won't open — copy the full link (it ends in a document ID); if it 404s, generate the document again.
• Still stuck? — open a GitHub issue or email [email protected].

License

MIT