Nutrient DWS MCP Server

Official MCP server for Nutrient Document Web Services: AI-driven PDF and Office document processing, conversion, OCR, redaction, and digital signing.

Developer Tools by Nutrient (PSPDFKit) API Key or OAuth2 active

GitHub Docs npm: @nutrient-sdk/dws-mcp-server

Overview

Nutrient DWS MCP Server is the official Model Context Protocol implementation from Nutrient (formerly PSPDFKit) that exposes the Nutrient Document Web Services Processor API to AI assistants. It lets agents create, edit, convert, sign, OCR, and redact PDFs and Office documents (DOCX, XLSX, PPTX) through natural language, with input and output files staying on the local filesystem.

The server bundles a handful of high-level tools that map onto the full DWS Processor API: a generic document_processor that handles conversion, OCR, watermarking, rotation, extraction, and redaction; a document_signer that applies CMS/PKCS#7 and CAdES (PAdES) signatures; an ai_redactor for detecting and removing PII like names, addresses, SSNs, and emails; plus utility tools for listing sandbox files and checking remaining DWS credits.

Notable features include an optional sandbox mode that restricts all file I/O to a configured directory, support for both API key and browser-based OAuth2 authentication, and stdio transport that works out of the box with Claude Desktop, Cursor, Windsurf, and VS Code (GitHub Copilot). The repo is maintained by Nutrient/PSPDFKit on the official org, making this the canonical integration.

Tools

Tool	Description
`document_processor`	Performs conversions between PDF, DOCX, images, PDF/A, PDF/UA, HTML, Markdown; merges documents; runs OCR; extracts text, tables, and key-value pairs; applies watermarks, page rotation, annotation flattening, and redaction.
`document_signer`	Applies digital signatures to PDFs using CMS/PKCS#7 or CAdES (PAdES-compliant), with support for visible or invisible signatures.
`ai_redactor`	Uses AI to detect and remove sensitive content (names, addresses, SSNs, emails, custom criteria) from documents.
`check_credits`	Read-only lookup of remaining DWS credits and account usage.
`sandbox_file_tree`	Read-only inspection of files within the configured sandbox directory.
`directory_tree`	Read-only view of local files when sandbox mode is disabled.

Setup Guide

Prerequisites

Node.js 18 or later
A free Nutrient account (sign up at nutrient.io/api)
A directory to use as the sandbox for file operations (strongly recommended)

Authentication

Two options:

API key: set NUTRIENT_DWS_API_KEY in the server env. Best for headless and CI use.
OAuth browser flow: omit NUTRIENT_DWS_API_KEY and the server will launch a browser consent flow on first call, caching tokens at ~/.config/nutrient/credentials.json.

Claude Desktop config

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "nutrient-dws": {
      "command": "npx",
      "args": ["-y", "@nutrient-sdk/dws-mcp-server"],
      "env": {
        "SANDBOX_PATH": "/your/sandbox/directory",
        "NUTRIENT_DWS_API_KEY": "your-api-key"
      }
    }
  }
}

Cursor config

In .cursor/mcp.json:

{
  "mcpServers": {
    "nutrient-dws": {
      "command": "npx",
      "args": ["-y", "@nutrient-sdk/dws-mcp-server"],
      "env": {
        "SANDBOX_PATH": "/your/project/documents"
      }
    }
  }
}

VS Code (GitHub Copilot)

In .vscode/mcp.json:

{
  "servers": {
    "nutrient-dws": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@nutrient-sdk/dws-mcp-server"],
      "env": {
        "SANDBOX_PATH": "${workspaceFolder}"
      }
    }
  }
}

Environment variables

NUTRIENT_DWS_API_KEY (optional): API key, otherwise OAuth is used
SANDBOX_PATH (recommended): restricts file I/O to this directory
AUTH_SERVER_URL (optional): defaults to https://api.nutrient.io
CLIENT_ID (optional): OAuth client ID for refresh token reuse
DWS_API_BASE_URL (optional): defaults to https://api.nutrient.io
LOG_LEVEL (optional): Winston log level, default info
MCP_LOG_FILE (optional): custom log file path

Restart your AI client after editing the config. Place documents in the sandbox directory and reference them by name in prompts.

Use Cases

Convert a batch of DOCX or image files into PDF/A archival format and OCR scanned pages into searchable text
Auto-redact PII (names, SSNs, emails, addresses) from contracts before sharing them with external parties
Merge multiple PDFs and Office documents into a single deliverable, then watermark and flatten annotations
Apply PAdES-compliant digital signatures (visible or invisible) to executed agreements straight from an agent workflow
Extract structured data (text, tables, key-value pairs) from invoices or forms into JSON for downstream automation

Example Prompts

"Merge invoice-1.pdf and invoice-2.pdf, add a 'PAID' watermark, and save as combined-invoices.pdf"
"Run OCR on scanned-contract.pdf and then redact every SSN and email address"
"Convert all DOCX files in the sandbox to PDF/A-2b"
"Extract the line item table from purchase-order.pdf as JSON"
"Apply an invisible PAdES signature to nda.pdf and check my remaining DWS credits"

Pros

Official server maintained by Nutrient/PSPDFKit on their GitHub org
Broad coverage of the DWS Processor API: conversion, OCR, redaction, signing, extraction, watermarking
Sandbox mode constrains file I/O for safer agent use
Supports both API key and OAuth2 browser auth flows

Limitations

Requires a Nutrient DWS account and consumes paid credits beyond the free tier
README states it is tested only on macOS and Windows (Linux support not officially claimed)
Only six high-level tools, so very specific DWS Processor operations must be expressed through prompt parameters rather than dedicated tool calls

Alternatives

Nutrient Document Engine MCP Server: self-hosted Nutrient Document Engine variant for the same use cases
mcp-pdf-tools: open-source local PDF manipulation server (no SaaS dependency)
Adobe PDF Services API (via custom MCP wrappers) for similar managed PDF processing