Documents MCP

An MCP (Model Context Protocol) server that provides AI agents with tools to create and read PDF, DOCX, and PPTX documents.

Features

• Create PDF - Generate PDF documents with text, headings, tables, images, and page numbers
• Create DOCX - Generate Word documents with headings, paragraphs, lists, tables, and images
• Create PPTX - Generate PowerPoint presentations with slides, text, shapes, tables, and charts
• Read PDF - Extract text content and metadata from PDF files (supports Gemini AI analysis)
• Read DOCX - Extract text and HTML content from Word documents (supports Gemini AI analysis)
• Read PPTX - Extract text content from PowerPoint presentations (supports Gemini AI analysis)

Installation

npm install documents-mcp

Or use directly with npx:

npx documents-mcp

Usage with Claude Desktop

Add to your Claude Desktop configuration (~/.config/claude/claude_desktop_config.json on macOS/Linux or %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "documents-mcp": {
      "command": "npx",
      "args": ["documents-mcp"],
      "env": {
        "OUTPUT_DIR": "/path/to/output/documents"
      }
    }
  }
}

Environment Variables

All environment variables are optional. Configure only the providers you need:

AI Provider API Keys

| Variable | Description | |----------|-------------| | OPENAI_API_KEY | OpenAI API key | | ANTHROPIC_API_KEY | Anthropic API key | | GOOGLE_API_KEY | Google AI API key | | GEMINI_API_KEY | Gemini API key (alias for GOOGLE_API_KEY) | | PERPLEXITY_API_KEY | Perplexity API key | | XAI_API_KEY | xAI (Grok) API key | | GROQ_API_KEY | Groq API key |

Local Model Endpoints

| Variable | Default | Description | |----------|---------|-------------| | OLLAMA_BASE_URL | http://localhost:11434 | Ollama server URL | | LMSTUDIO_BASE_URL | http://localhost:1234 | LM Studio server URL | | VLLM_BASE_URL | http://localhost:8000 | vLLM server URL |

Output Configuration

| Variable | Default | Description | |----------|---------|-------------| | OUTPUT_DIR | Current directory | Default directory for saved documents |

Tool Reference

create-pdf

Create a PDF document with structured content.

Parameters:

• title (required): Document title
• author (optional): Document author
• content (required): Array of content items:
  • { type: "text", content: string, fontSize?: number, bold?: boolean, color?: {r, g, b} }
  • { type: "heading", content: string, level: 1-6 }
  • { type: "table", headers: string[], rows: string[][] }
  • { type: "image", base64: string, format: "png"|"jpg"|"jpeg", width?: number, height?: number }
  • { type: "pageBreak" }
• outputPath (optional): File path to save the PDF
• pageSize (optional): "A4", "Letter", or "Legal"

create-docx

Create a Word document with rich formatting.

Parameters:

• title (required): Document title
• author (optional): Document author
• content (required): Array of content items:
  • { type: "text", content: string, bold?: boolean, italic?: boolean, underline?: boolean }
  • { type: "heading", content: string, level: 1-6 }
  • { type: "paragraph", content: string, alignment?: "left"|"center"|"right"|"justified" }
  • { type: "bulletList", items: string[] }
  • { type: "numberedList", items: string[] }
  • { type: "table", headers: string[], rows: string[][] }
  • { type: "image", base64: string, width?: number, height?: number }
  • { type: "pageBreak" }
• outputPath (optional): File path to save the DOCX

create-pptx

Create a PowerPoint presentation.

Parameters:

• title (required): Presentation title
• author (optional): Presentation author
• slides (required): Array of slide objects:
  • title (optional): Slide title
  • subtitle (optional): Slide subtitle
  • layout: "title", "titleAndContent", "blank", or "sectionHeader"
  • elements: Array of elements:
    • { type: "textBox", text: string, x?, y?, w?, h?, fontSize?, bold?, color?, align? }
    • { type: "image", base64: string, x?, y?, w?, h? }
    • { type: "shape", shapeType: "rect"|"ellipse"|"triangle"|"line"|"arrow", ... }
    • { type: "table", headers: string[], rows: string[][], x?, y?, w? }
    • { type: "chart", chartType: "bar"|"line"|"pie"|"doughnut", data: [...], ... }
  • backgroundColor (optional): Slide background color
  • notes (optional): Speaker notes
• outputPath (optional): File path to save the PPTX

read-pdf

Extract text content from a PDF file.

Parameters:

• filePath (optional): Path to the PDF file
• base64Content (optional): Base64-encoded PDF content
• prompt (optional): Instruction for AI analysis (requires GOOGLE_API_KEY)

Returns: { text, metadata: { pageCount, info, version }, characterCount, wordCount, aiAnalysis? }

read-docx

Extract text content from a Word document.

Parameters:

• filePath (optional): Path to the DOCX file
• base64Content (optional): Base64-encoded DOCX content
• outputFormat (optional): "text", "html", or "both"
• prompt (optional): Instruction for AI analysis (requires GOOGLE_API_KEY)

Returns: { text?, html?, characterCount, wordCount, aiAnalysis? }

read-pptx

Extract text content from a PowerPoint presentation.

Parameters:

• filePath (optional): Path to the PPTX file
• base64Content (optional): Base64-encoded PPTX content
• prompt (optional): Instruction for AI analysis (requires GOOGLE_API_KEY)

Returns: { text, slideCount, slides: [{slide, text}], aiAnalysis? }

Running as HTTP Server

Start the HTTP server for web-based MCP clients:

npm run start:http
# or
npx documents-mcp-http

The HTTP server exposes:

• GET /sse - SSE endpoint for MCP clients
• POST /messages?sessionId=<id> - Message endpoint for SSE sessions
• GET /health - Health check endpoint

Client SDK

Use the client SDK to connect to the documents-mcp server programmatically:

import { createClient } from "documents-mcp/client";

// Connect via STDIO (for CLI usage)
const client = createClient({
  transport: "stdio",
  command: "documents-mcp",
});

await client.connect();

// List available tools
const tools = await client.listTools();
console.log(tools);

// Create a PDF
const result = await client.createPdf({
  title: "My Document",
  content: [
    { type: "heading", content: "Hello World", level: 1 },
    { type: "text", content: "This is a sample document." },
  ],
});

console.log(result.content); // { success: true, filePath: "...", pageCount: 1 }

await client.disconnect();

SSE Transport

For HTTP/SSE connections:

const client = createClient({
  transport: "sse",
  url: "http://localhost:3000/sse",
});

await client.connect();
// ... use client methods
await client.disconnect();

Available Client Methods

| Method | Description | |--------|-------------| | connect() | Connect to the MCP server | | disconnect() | Disconnect from the server | | listTools() | List available tools | | callTool(name, args) | Call any tool by name | | createPdf(options) | Create a PDF document | | createDocx(options) | Create a Word document | | createPptx(options) | Create a PowerPoint presentation | | readPdf(options) | Extract text from a PDF | | readDocx(options) | Extract text from a Word document | | readPptx(options) | Extract text from a PowerPoint |

Programmatic Usage (Direct)

You can also use the document tools directly without the MCP server:

import { createPdf, readPdf } from "documents-mcp";

// Create a PDF
const result = await createPdf.handler(
  createPdf.schema.parse({
    title: "My Document",
    content: [{ type: "text", content: "Hello World" }],
  })
);

// Read a PDF
const extracted = await readPdf.handler(
  readPdf.schema.parse({
    filePath: "/path/to/document.pdf",
  })
);

Development

# Clone the repository
git clone https://github.com/HarjjotSinghh/documents-mcp.git
cd documents-mcp

# Install dependencies
npm install

# Run in development mode
npm run dev

# Build for production
npm run build

License

MIT