File Converter MCP Server

简体中文 | English

This MCP server provides multiple file conversion tools for converting various document and image formats. This project is built using the Model Context Protocol (MCP) and is designed to serve AI agents that need file conversion capabilities.

Features

• DOCX to PDF: Convert Microsoft Word documents to PDF
• PDF to DOCX: Convert PDF documents to Microsoft Word format
• Image Format Conversion: Convert between various image formats (JPG, PNG, WebP, etc.)
• Excel to CSV: Convert Excel spreadsheets to CSV format
• HTML to PDF: Convert HTML files to PDF format
• Markdown to PDF: Convert Markdown documents to PDF with proper styling
• Generic Conversion: A versatile tool that attempts to handle various format conversions

Technologies

• Python 3.10+
• Model Context Protocol (MCP) Python SDK
• Various conversion libraries:
  • docx2pdf - for DOCX to PDF conversion
  • pdf2docx - for PDF to DOCX conversion
  • Pillow - for image format conversions
  • pandas - for Excel to CSV conversion
  • pdfkit - for HTML to PDF conversion
  • markdown - for Markdown to HTML conversion

Installation

1. Clone the Repository

   git clone https://github.com/wowyuarm/file-converter-mcp.git
   cd file-converter-mcp
   

2. Create a Virtual Environment (optional but recommended)

   python -m venv venv
   source venv/bin/activate      # On Unix-based systems
   venv\Scripts\activate         # On Windows
   

3. Install Dependencies

   Install the required packages using pip:

   pip install mcp docx2pdf pdf2docx pillow pandas pdfkit markdown
   

   Alternatively, if you are using uv:

   uv add "mcp[cli]" docx2pdf pdf2docx pillow pandas pdfkit markdown
   

   Note: Some conversion libraries may have additional system dependencies. Please check their documentation for details.

Usage

Running the Server in Development Mode

To test the server, run:

mcp dev file_converter_server.py

Installing for Claude Desktop

Optionally, you can install the server on Claude Desktop with:

mcp install file_converter_server.py --name "File Converter"

API / Tools

The MCP server exposes the following tools:

Path-Based Tools (Also Support Content Input)

docx2pdf

Command: docx2pdf

• Input Option 1: Path to a .docx file

  input_file: path/to/document.docx
  

• Input Option 2: Base64 encoded content of the DOCX file

  file_content_base64: [base64 encoded string]
  

• Output: Base64 encoded string of the converted PDF file

pdf2docx

Command: pdf2docx

• Input Option 1: Path to a PDF file

  input_file: path/to/document.pdf
  

• Input Option 2: Base64 encoded content of the PDF file

  file_content_base64: [base64 encoded string]
  

• Output: Base64 encoded string of the converted DOCX file

convert_image

Command: convert_image

• Input Option 1:

  input_file: path/to/image.png
  output_format: jpg
  

• Input Option 2:

  file_content_base64: [base64 encoded string]
  input_format: png
  output_format: jpg
  

• Output: Base64 encoded string of the converted image

excel2csv

Command: excel2csv

• Input: Path to an Excel file (.xls or .xlsx)
• Output: Base64 encoded string of the converted CSV file

html2pdf

Command: html2pdf

• Input: Path to an HTML or Markdown file (.html, .md, .markdown)
• Output: Base64 encoded string of the converted PDF file

convert_file (Generic Converter)

Command: convert_file

• Input Option 1:

  input_file: path/to/file.docx
  input_format: docx
  output_format: pdf
  

• Input Option 2:

  file_content_base64: [base64 encoded string]
  input_format: docx
  output_format: pdf
  

• Output: Base64 encoded string of the converted file

Content-Based Tools (Legacy)

These are maintained for backward compatibility. All main tools now support content-based input directly.

convert_content (Generic Content Converter)

Command: convert_content

• Input:
  • Base64 encoded content of the input file
  • Source format (e.g., "docx", "pdf", "md")
  • Target format (e.g., "pdf", "docx")
• Output: Base64 encoded string of the converted file

docx2pdf_content

Command: docx2pdf_content

• Input: Base64 encoded content of the DOCX file
• Output: Base64 encoded string of the converted PDF file

pdf2docx_content

Command: pdf2docx_content

• Input: Base64 encoded content of the PDF file
• Output: Base64 encoded string of the converted DOCX file

markdown2pdf_content

Command: markdown2pdf_content

• Input: Base64 encoded content of the Markdown file
• Output: Base64 encoded string of the converted PDF file

File Handling

The server includes robust file path handling that:

• Uses a multi-stage search strategy to find files
• Searches for uploaded files in common locations (temp directories, current directory)
• Tries multiple filename variations (case-insensitive, with/without extensions)
• Provides detailed logs to help troubleshoot file location issues
• Works seamlessly with files uploaded via Claude chat interface
• Supports relative and absolute file paths
• Automatically detects file formats when possible

Dual-Mode Input

All conversion tools now support two methods of input:

1. Path-Based Conversion (traditional approach)

   @File Converter
   docx2pdf
   input_file: file.docx
   

2. Content-Based Conversion (works even when path lookup fails)

   @File Converter
   docx2pdf
   file_content_base64: [base64 encoded string]
   

This dual-mode approach provides maximum flexibility and reliability:

• When in doubt, use content-based input for guaranteed processing
• All intermediate files are created with unique names in temporary directories
• Temporary files are automatically cleaned up after processing

Handling Claude-Specific File Uploads

When using with Claude, if a file upload fails to be found:

1. Try using the original filename with a preceding path:

   @File Converter
   docx2pdf
   input_file: /tmp/file.docx
   

2. If that fails, obtain the file content directly from Claude:

   @File Converter
   docx2pdf
   file_content_base64: [base64 content obtained from Claude]
   

Error Handling

• Each tool validates file existence using multiple search strategies
• Detailed error messages are returned in a structured JSON format: {"success": false, "error": "error message"}
• Successful conversions return: {"success": true, "data": "base64 encoded file content"}
• The server includes comprehensive logging for troubleshooting
• The server gracefully handles exceptions and returns informative error messages

Contributing

Contributions are welcome! If you'd like to contribute, please follow the guidelines in CONTRIBUTING.md (中文版: 贡献指南, English: Contributing Guidelines).

License

This project is licensed under the MIT License – see the LICENSE file for details.

GitHub Repository

Visit the GitHub repository at: https://github.com/wowyuarm/file-converter-mcp