polydoc-mcp

An MCP server for PolyDoc, a REST API that converts HTML or URLs to PDF, captures screenshots, and generates EU-compliant e-invoices (Factur-X / ZUGFeRD hybrid PDF/A-3). It lets MCP clients (Claude Desktop, Claude Code, Cursor, and others) drive PolyDoc directly.

Tools

• polydoc_html_to_pdf - HTML, URL, or saved template to PDF (layout, margins, page format, page ranges, bookmarks, accessible/tagged PDFs).

• polydoc_screenshot - HTML, URL, or template to PNG / JPEG / WebP, with viewport and device-pixel-ratio control. Returns an inline image preview when small enough.

• polydoc_generate_einvoice - Factur-X or ZUGFeRD hybrid PDF/A-3 from structured invoice data, profiles from minimum to extended.

• polydoc_test_credentials - verify the configured API key with a minimal sandbox render (never draws production quota).

Content can come from a URL, an inline HTML string, or a saved template (with Liquid templateData). Downloads are written to a local directory and the file path is returned; you can also deliver to your cloud storage (presigned URL) or a webhook.

Related MCP server: facturx-mcp

Installation

Requires Node.js >= 20.15. Published to npm, so clients can run it with npx:

npx -y polydoc-mcp

The server is configured entirely through environment variables (see below).

Configuration

Client setup

Claude Desktop

Add to claude_desktop_config.json (Settings > Developer > Edit Config):

{
  "mcpServers": {
    "polydoc": {
      "command": "npx",
      "args": ["-y", "polydoc-mcp"],
      "env": {
        "POLYDOC_API_KEY": "your-key",
        "POLYDOC_OUTPUT_DIR": "/Users/you/Documents/polydoc"
      }
    }
  }
}

Claude Code

claude mcp add polydoc -e POLYDOC_API_KEY=your-key -- npx -y polydoc-mcp

Cursor

Add to ~/.cursor/mcp.json (or a project .cursor/mcp.json) using the same mcpServers block shown for Claude Desktop.

Output and delivery

• Download (default): the file is written under POLYDOC_OUTPUT_DIR and the absolute path is returned with metadata (size, conversion id, credits). Screenshots also return an inline image block. Pass returnBase64: true to also receive the bytes as base64 (heavy; for clients without a filesystem).

• Cloud storage: set delivery: "cloudStorage" and presignedUrl. The upload URL is returned.

• Webhook: set delivery: "webhook" and a webhook object ({ url, async?, method?, headers? }).

Anything not in a tool's schema?

Every conversion tool has an advanced object that is deep-merged into the request body, so any API capability not surfaced as a typed field (for example pdf.watermark, pdf.encryption, render, request) is still reachable. See the field reference at docs.polydoc.tech.

Examples

Worked tool inputs for each angle live in examples/: a branded PDF from a template, a URL screenshot, and a ZUGFeRD / EN 16931 e-invoice.

Development

• yarn install

• yarn build - compile to dist/

• yarn lint

• yarn test - unit tests (request builder, output jail, tool handlers) and an in-memory MCP round-trip

• POLYDOC_API_KEY=your-key yarn test:integration - live smoke tests against the sandbox

• yarn scrub:check - check for em-dashes in source and docs

Run the built server directly for a quick check:

POLYDOC_API_KEY=your-key node dist/index.js

License

MIT