Document Loader MCP Server

Model Context Protocol (MCP) server for document parsing and content extraction

This MCP server provides tools to parse and extract content from various document formats including PDF, Word documents, Excel spreadsheets, PowerPoint presentations, and images.

Features

• PDF Text Extraction: Extract text content from PDF files using pdfplumber
• Word Document Processing: Convert DOCX/DOC files to markdown using markitdown
• Excel Spreadsheet Reading: Parse XLSX/XLS files and convert to markdown
• PowerPoint Presentation Processing: Extract content from PPTX/PPT files
• Image Loading: Load and display various image formats (PNG, JPG, GIF, BMP, TIFF, WEBP)
• Slide Image Extraction: Extract individual slides/pages as PNG images from PPTX, PPT, or PDF files using LibreOffice and poppler

Prerequisites

Installation Requirements

1. Install uv from Astral or the GitHub README
2. Install Python 3.10 or newer using uv python install 3.10 (or a more recent version)

Optional: Slide Image Extraction

The extract_slides_as_images tool requires external system packages:

• LibreOffice (for PPTX/PPT → PDF conversion):
  • Ubuntu/Debian: sudo apt install libreoffice
  • macOS: brew install --cask libreoffice
  • Windows: Download from libreoffice.org
• poppler-utils (for PDF → image rendering):
  • Ubuntu/Debian: sudo apt install poppler-utils
  • macOS: brew install poppler
  • Windows: Download from GitHub and add to PATH

Installation

Kiro	Cursor	VS Code

Configure the MCP server in your MCP client configuration:

{
  "mcpServers": {
    "awslabs.document-loader-mcp-server": {
      "command": "uvx",
      "args": ["awslabs.document-loader-mcp-server@latest"],
      "env": {
        "FASTMCP_LOG_LEVEL": "ERROR"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

For Kiro MCP configuration, see the Kiro IDE documentation or the Kiro CLI documentation for details.

For global configuration, edit ~/.kiro/settings/mcp.json. For project-specific configuration, edit .kiro/settings/mcp.json in your project directory.

Available Tools

• read_document: Extract content from various document formats by specifying file_path and file_type ('pdf', 'docx', 'doc', 'xlsx', 'xls', 'pptx', 'ppt')
• read_image: Load image files for LLM viewing and analysis
• extract_slides_as_images: Extract slides/pages as individual PNG images from PPTX, PPT, or PDF files. Requires LibreOffice (for PPTX/PPT) and poppler-utils (for PDF-to-image rendering)

Environment Variables

• FASTMCP_LOG_LEVEL: Set logging level (ERROR, INFO, DEBUG)
• MAX_FILE_SIZE_MB: Maximum allowed file size in megabytes (default: 50). Must be a positive integer.
• DOCUMENT_BASE_DIR: Base directory that constrains all file access (read_document, read_image, and both the input and output paths of extract_slides_as_images). Every resolved path must be inside this directory or the request is denied. Defaults to the current working directory. Set it explicitly (e.g. /var/app/documents) to define your sandbox. The sandbox is always enforced — there is no automatic bypass, including under CI (CI, GITHUB_ACTIONS) or test (PYTEST_CURRENT_TEST) environments.

Development

Setup

# Clone the repository
git clone https://github.com/awslabs/mcp.git
cd mcp/src/document-loader-mcp-server

# Install dependencies
uv sync

# Install in development mode
uv pip install -e .

Testing

# Run tests
uv run pytest

# Run with coverage
uv run pytest --cov=awslabs.document_loader_mcp_server

The test suite includes:

• Server functionality validation
• Document parsing tests with generated sample files
• Error handling verification

Sample Documents

The test suite automatically generates sample documents for testing:

• PDF with multi-page content
• DOCX with formatted text and lists
• XLSX with multiple sheets and data
• PPTX with slides and content
• Various image formats

Docker

You can also run this server in a Docker container. The image speaks MCP over stdio (the Dockerfile does not expose an HTTP port), so use --interactive and omit -p. Mount a host docs folder so tools can see files, and set DOCUMENT_BASE_DIR to that mount (paths you pass to tools are container paths like /docs/report.pdf). Add :ro to the mount if you only read documents; leave it writable for extract_slides_as_images.

docker build -t awslabs/document-loader-mcp-server .
docker run --rm --interactive \
  -v "$PWD/docs:/docs" \
  -e DOCUMENT_BASE_DIR=/docs \
  -e FASTMCP_LOG_LEVEL=ERROR \
  awslabs/document-loader-mcp-server

Example MCP client config:

{
  "mcpServers": {
    "awslabs.document-loader-mcp-server": {
      "command": "docker",
      "args": [
        "run", "--rm", "--interactive",
        "-v", "/path/to/docs:/docs",
        "-e", "DOCUMENT_BASE_DIR=/docs",
        "-e", "FASTMCP_LOG_LEVEL=ERROR",
        "awslabs/document-loader-mcp-server"
      ]
    }
  }
}

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Contributing

We welcome contributions! Please see CONTRIBUTING.md for details.

Support

For issues and questions, please use the GitHub issue tracker.