Overview

The Nutrient MCP Server is an implementation of the Model Context Protocol (MCP) that helps developers build document workflows driven by natural language. It connects a language model with document tooling (PDF operations, OCR, signing) so you can create, edit, sign and extract data from documents using plain-English prompts rather than wiring low-level PDF APIs yourself.

Designed for integration into backend services, the server acts as a translator between natural-language instructions and concrete document operations. Typical uses include converting prompts into filled forms, extracting structured fields from scanned documents, programmatically applying digital signatures, and generating or modifying contract text — all while keeping the document assets and signing credentials under your control.

Features

Create documents from templates or from scratch using natural language prompts
Edit document content (text replacement, field insertion, redaction) via instructions
Extract structured data (form fields, tables, named entities) from PDFs/OCR results
Apply digital signatures and prepare documents for e-sign workflows
Pluggable model/back-end providers (LLMs) and configurable document processors
REST API for easy integration with backends, server-side automation, or job queues
Local or cloud storage support and audit-friendly operation logs

Installation / Configuration

Prerequisites: Node.js (>= 16) or Docker, and access credentials for any integrated services (LLM provider, PSPDFKit or other PDF processor, signing backend).

Clone the repository and install dependencies:

git clone https://github.com/PSPDFKit/nutrient-dws-mcp-server.git
cd nutrient-dws-mcp-server
npm install

Create a copy of the example environment file and edit configuration values:

cp .env.example .env
# then open .env and set the variables

Example .env (adjust to your environment):

PORT=3000
MCP_MODEL=openai-gpt-4o
MODEL_API_KEY=sk-...
PSPDFKIT_SERVER_URL=https://your-pspdfkit-server
PSPDFKIT_API_KEY=pspdfkit_...
STORAGE_DIR=./storage
SIGNING_KEY_PATH=/secrets/signing-key.pem
LOG_LEVEL=info

Typical npm scripts:

# start in production
npm start

# start in development with auto-reload
npm run dev

Docker

# build
docker build -t nutrient-mcp-server .

# run (example)
docker run -d \
  -p 3000:3000 \
  -e PORT=3000 \
  -e MODEL_API_KEY=sk-... \
  -e PSPDFKIT_API_KEY=pspdfkit_... \
  -v /local/storage:/app/storage \
  nutrient-mcp-server

Configuration reference (common environment variables):

Variable	Description
PORT	HTTP port to listen on
MCP_MODEL	Identifier for the model or connector to use
MODEL_API_KEY	API key for your LLM provider
PSPDFKIT_SERVER_URL	Optional URL for a PSPDFKit backend
PSPDFKIT_API_KEY	API key for PSPDFKit operations
STORAGE_DIR	Local directory or mount for uploaded documents
SIGNING_KEY_PATH	Path to private key or signing credential store
LOG_LEVEL	Logging verbosity (error, warn, info, debug)

Available Tools / Resources

GitHub repository: https://github.com/PSPDFKit/nutrient-dws-mcp-server
Built-in API documentation: the server typically exposes an OpenAPI/Swagger spec at /openapi.json (check your running instance)
Example scripts: look for an examples/ or clients/ folder in the repo showing sample requests and response handling
Integrations: adapters for PDF processing (PSPDFKit or other), OCR modules, and signing backends (configurable)

Typical request format

The server expects an MCP-style JSON payload describing the intent, document(s), and contextual metadata. A minimal example:

curl -X POST http://localhost:3000/api/mcp \
  -H "Authorization: Bearer <YOUR_TOKEN>" \
  -H "Content-Type: application/json" \
  -d '{
    "action": "edit",
    "instructions": "Insert a signature field at bottom of page 2 and replace [COMPANY] with \"Acme Inc.\"",
    "documentUrl": "https://example.com/contracts/draft.pdf",
    "metadata": { "requestId": "1234" }
  }'

Common action values: create, edit, sign, extract. The server returns a job result with operation status, a link to the modified document, and structured outputs when applicable (extracted fields, audit log).

Use Cases

Form extraction and onboarding: Upload scanned onboarding forms and use a single prompt like “Extract name, address, email, and employer from this page” to populate your CRM with structured JSON.
Contract drafting and edits: Start from a short brief — “Create a one-page contractor agreement for freelance designers with payment terms and a 14-day termination clause” — and receive a PDF draft you can further refine or send for signature.
Bulk signing automation: Programmatically attach signature fields and apply signatures for batches of

Nutrient MCP Server: Create Edit Sign Extract Documents