PDFGate MCP Server

What PDFGate MCP Server does

PDFGate MCP Server exposes PDFGate operations as MCP tools for AI assistants and other MCP clients. It covers the document lifecycle from generating a PDF from a URL or HTML through uploading, retrieving, downloading, transforming, protecting, watermarking, compressing, and deleting stored documents.

The server also provides e-signature workflows. Agents can assemble envelopes from existing documents, send them to recipients, inspect signing progress, cancel or delete envelopes, and create short-lived embedded signing links for use inside an application. Recipient records can be created and reused by ID, and webhook subscriptions can be managed through MCP.

How it works

A local deployment runs through the npm package with npx and communicates with PDFGate using an API key. PDFGate keys beginning with live_ target production, while keys beginning with test_ select the sandbox. The hosted deployment is available at https://mcp.pdfgate.com/server and supports Streamable HTTP. Clients can authenticate there with OAuth or by sending the X-API-KEY header.

Most document-processing operations work with stored PDFGate document IDs. For example, an agent can upload a PDF, add fields or apply protection, and then use the resulting document in another operation. Transformations such as flattening and encryption create a new document instead of changing the original.

Envelope creation accepts one or more existing documents. Recipients may be supplied directly by name and email or referenced by stored recipient ID. Ordered signing can activate recipients in sequence. Embedded recipients do not receive email; after an envelope is sent, the client can request a signing URL that expires after 10 minutes.

Setup and configuration

For local execution, add a server entry similar to this to the MCP client configuration:

{
  "mcpServers": {
    "pdfgate": {
      "command": "npx",
      "args": ["-y", "@pdfgate/mcp-server"],
      "env": {
        "PDFGATE_API_KEY": "your_api_key_here"
      }
    }
  }
}

The hosted endpoint can be configured as a remote HTTP MCP server. OAuth is suitable for clients that support remote MCP authentication; API-key clients can pass X-API-KEY in the request headers. The exact configuration property names vary by client.

Webhook delivery uses port 3599 by default. To receive events, expose the listener publicly and register the public URL with create_webhook. The port can be changed with PDFGATE_WEBHOOK_PORT.

Tools and capabilities

• Generate PDFs from a URL or raw HTML, and upload local files or remote URLs.
• Retrieve document metadata, download PDF bytes, or delete stored documents.
• Flatten interactive PDFs, extract form values, and add fields using placeholders or explicit positions.
• Compress, password-protect, and watermark PDFs with text or images.
• Create, send, inspect, void, and delete signing envelopes.
• Create embedded signing links and manage stored recipients.
• Create, inspect, and delete webhook subscriptions.

Supported webhook events include envelope delivery, completion, expiration, voiding, deletion, document completion, recipient signing, and recipient signature changes. Incoming events are available through pdfgate://events.

Limitations and notes

generate_pdf requires either url or html, not both. Likewise, upload_file accepts either a local filePath or a URL. Embedded signing links are short-lived and should be created when the signer is ready. For ordered envelopes, a recipient must be active before a link can be created.

Voiding an active envelope cannot be undone. Deleting an envelope permanently removes its produced signed documents and audit logs, while source documents remain available. Webhook signing secrets returned during creation should be stored securely so incoming event signatures can be verified.