Pandoc MCP Server

A Python-based MCP (Model Context Protocol) server that provides powerful document conversion capabilities via Pandoc. This server allows AI agents (like Claude via LangChain/LangGraph) to request file conversions between various formats such as Markdown, DOCX, HTML, PDF, EPUB, and many more.

This project uses:

• FastMCP: A Python library for easily creating MCP servers.
• pypandoc: A Python wrapper around the Pandoc command-line tool.
• Pandoc: The universal document converter.
• (Optional) Docker: For containerized deployment, bundling all dependencies (Python, Pandoc, LaTeX).

Features

• Exposes a single MCP tool: convert_document.
• Supports a wide range of input and output formats handled by Pandoc.
• Allows specifying input format (if auto-detection fails) and output format.
• Supports passing extra command-line arguments to Pandoc for advanced control (e.g., Table of Contents, PDF margins, standalone files).
• Includes Docker configuration (Dockerfile) for creating a self-contained server environment including Pandoc and necessary LaTeX components for PDF generation.
• Designed for integration with MCP clients, particularly LangChain/LangGraph agents.

Exposed MCP Tool

convert_document

Converts a document from one format to another using Pandoc.

Arguments:

• input_file_path (str, required): The path accessible by the server to the input document file. If running in Docker with a volume mount, this should be the path inside the container (e.g., /data/my_doc.docx).
• output_file_path (str, required): The path accessible by the server where the converted output file should be saved. If running in Docker, this should be the path inside the container (e.g., /data/my_output.pdf). The directory will be created if it doesn't exist within the server's accessible filesystem.
• to_format (str, required): The target format for the conversion (e.g., 'markdown', 'docx', 'pdf', 'html', 'rst', 'epub'). See Pandoc documentation for a full list (--list-output-formats).
• from_format (str, optional): The format of the input file. If None, pandoc will try to guess from the file extension. Specify if the extension is ambiguous or missing (e.g., 'md', 'docx', 'html'). Defaults to None.
• extra_args (List[str], optional): A list of additional command-line arguments to pass directly to pandoc (e.g., ['--toc'], ['-V', 'geometry:margin=1.5cm'], ['--standalone']). Defaults to None.

Returns:

• (str): A message indicating success (e.g., "Successfully converted document to '/data/my_output.pdf'") or an error message (e.g., "Error: Input file not found...", "Error during conversion: Pandoc died...").

Setup and Running

You can run this server either locally (requires manual installation of dependencies) or using the provided Docker configuration (recommended for ease of use and deployment).

Installing via Smithery

To install Pandoc Document Converter for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install @MaitreyaM/file-converter-mcp --client claude

Option 1: Running with Docker (Recommended)

This method bundles Python, Pandoc, LaTeX, and required libraries into a container. You only need Docker Desktop installed locally.

1. Install Docker: Download and install Docker Desktop for your operating system. Start Docker Desktop.
2. Clone Repository: Get the project files:

   git clone https://github.com/your-username/pandoc-mcp-server.git # Replace with your repo URL
   cd pandoc-mcp-server
   

3. Build the Docker Image: This command builds the image using the Dockerfile. It installs Pandoc, a capable TeX Live distribution (for PDF support), and Python dependencies inside the image. This step might take several minutes the first time.

   docker build -t pandoc-converter-server .
   

4. Run the Container: This starts the server inside the container.
   • Choose a directory on your host machine to share with the container for input/output files (e.g., the current project directory).
   • Run the container, mapping the host directory to /data inside the container and mapping port 8000. Replace /path/to/your/local/project with the actual absolute path to the project directory on your machine.

   # Example using the current directory (.) as the host path:
   docker run -it --rm -p 8000:8000 -v "$(pwd)":/data pandoc-converter-server
   
   # Or using an absolute path (replace):
   # docker run -it --rm -p 8000:8000 -v "/path/to/your/local/project":/data pandoc-converter-server
   

   • -it: Runs interactively (shows logs, allows Ctrl+C).
   • --rm: Removes the container when stopped.
   • -p 8000:8000: Maps port 8000 on your host to port 8000 in the container.
   • -v "$(pwd)":/data: Mounts the current working directory on your host to /data inside the container. Files placed in your local project directory will appear in /data inside the container, and files saved to /data by the server will appear in your local project directory.
   • pandoc-converter-server: The name of the image you built.
5. Server is Running: You should see logs indicating the server started and is listening on SSE (http://0.0.0.0:8000). It's ready to accept connections from your MCP client (like the LangChain agent).
6. Connecting from Client: Configure your MCP client (e.g., MultiServerMCPClient) to connect to http://127.0.0.1:8000/sse with transport: "sse".
7. Using the Tool: When interacting with your agent/client, refer to files using their path inside the container, prefixed with /data/. For example: convert /data/my_input.docx to pdf at /data/my_output.pdf. The output file will appear in your local project directory due to the volume mapping.

Option 2: Running Locally (Manual Dependency Installation)

This requires you to install Python, Pandoc, and a LaTeX distribution directly onto your host machine.

1. Install Python: Ensure you have Python >= 3.10 installed.
2. Install Pandoc: Install the Pandoc command-line tool for your OS. Follow instructions at pandoc.org/installing.html. Verify by running pandoc --version in a new terminal.
3. Install LaTeX: For PDF generation, install a TeX distribution.
   • macOS: brew install --cask mactex-no-gui (Recommended via Homebrew)
   • Debian/Ubuntu: sudo apt-get update && sudo apt-get install texlive-latex-base texlive-fonts-recommended texlive-latex-extra texlive-fonts-extra (or texlive-full for everything, but large).
   • Windows: Install MiKTeX or TeX Live. Ensure the bin directory containing pdflatex.exe is added to your system's PATH.
   • Verify by running pdflatex --version in a new terminal.
4. Clone Repository:

   git clone https://github.com/your-username/pandoc-mcp-server.git # Replace with your repo URL
   cd pandoc-mcp-server
   

5. Create Virtual Environment (Recommended):

   python -m venv venv
   source venv/bin/activate # Linux/macOS
   # venv\Scripts\activate # Windows
   

   (Or use Conda: conda create --name pandoc-env python=3.11 && conda activate pandoc-env)
6. Install Python Dependencies:

   pip install -r requirements.txt
   

7. Run the Server:

   python pandoc_mcp_server.py
   

8. Server is Running: It will listen on http://127.0.0.1:8000/sse.
9. Connecting from Client: Configure your MCP client to connect to http://127.0.0.1:8000/sse.
10. Using the Tool: Refer to files using their regular paths on your local machine (e.g., convert my_input.docx to pdf at my_output.pdf, assuming files are in the same directory, or use absolute paths).

Example Agent Interaction (Running Server in Docker)

Assuming the server container is running with the volume mount:

You: convert /data/report.md to pdf

Agent: Thinking...
[Agent calls convert_document tool with input='/data/report.md', output='/data/report.pdf', to='pdf']
Agent: Successfully converted document to '/data/report.pdf'
[The bot may then attempt to upload report.pdf from the local project directory]

Files

• pandoc_mcp_server.py: The main Python script for the MCP server.
• Dockerfile: Instructions for building the Docker container image.
• requirements.txt: Python dependencies needed inside the Docker container (or local venv).
• .gitignore: Specifies intentionally untracked files for Git.
• README.md: This file.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request or open an Issue.