Mcp Document Converter

MCP Document Converter

MCP (Model Context Protocol) Document Converter - A powerful MCP tool for converting documents between multiple formats, enabling AI agents to easily transform documents.

🌐 Language: English | 中文

Features

• Multi-format Support: Supports 5 mainstream document formats: Markdown, HTML, DOCX, PDF, and Text
• Bidirectional Conversion: Any format can be converted to any other format (5×5=25 conversion combinations)
• MCP Protocol: Compliant with MCP standards, can be used as a tool for AI assistants like Trae IDE
• Plugin Architecture: Easy to extend with new parsers and renderers
• Syntax Highlighting: HTML and PDF outputs support code syntax highlighting
• Style Customization: Support for custom CSS styles
• Metadata Preservation: Preserves document title, author, creation time, and other metadata during conversion

---

📚 Documentation

[User Guide](docs/en/USER_GUIDE.md) · [API Reference](docs/en/API.md) · [Contributing](docs/en/CONTRIBUTING.md) · [Changelog](docs/en/CHANGELOG.md) · [License](LICENSE)

---

Architecture

flowchart TB
    subgraph Parsers["Parsers"]
        MD[Markdown]
        DOCX1[DOCX]
        HTML1[HTML]
        PDF1[PDF]
        TXT1[Text]
    end

    subgraph IR["Intermediate Representation (IR)"]
        DT[Document Tree]
        META[Metadata]
        ASSETS[Assets]
    end

    subgraph Renderers["Renderers"]
        HTML2[HTML]
        PDF2[PDF]
        MD2[Markdown]
        DOCX2[DOCX]
        TXT2[Text]
    end

    MD --> IR
    DOCX1 --> IR
    HTML1 --> IR
    PDF1 --> IR
    TXT1 --> IR
    
    IR --> HTML2
    IR --> PDF2
    IR --> MD2
    IR --> DOCX2
    IR --> TXT2

Core Components

1. DocumentIR (Intermediate Representation): Unified abstraction for all documents, containing document tree, metadata, assets, etc.
2. BaseParser (Parser Base Class): Defines the parser interface, parses various formats into DocumentIR
3. BaseRenderer (Renderer Base Class): Defines the renderer interface, renders DocumentIR into various formats
4. ConverterRegistry (Registry): Manages all parsers and renderers, provides format lookup and auto-matching
5. DocumentConverter (Conversion Engine): Coordinates parsers and renderers to complete document conversion

Supported Formats

Input Formats (Parsers)

| Format | Extensions | MIME Type | Features | |--------|------------|-----------|----------| | Markdown | .md, .markdown, .mdown, .mkd | text/markdown | YAML Front Matter, GFM extensions | | HTML | .html, .htm | text/html | Semantic tag parsing | | DOCX | .docx | application/vnd.openxmlformats-officedocument.wordprocessingml.document | Styles, tables, images | | PDF | .pdf | application/pdf | Text extraction and structure recognition | | Text | .txt, .text | text/plain | Auto encoding detection and structure recognition |

Output Formats (Renderers)

| Format | Extension | MIME Type | Features | |--------|-----------|-----------|----------| | HTML | .html | text/html | Beautiful styling, code highlighting, responsive design | | Markdown | .md | text/markdown | Standard Markdown format, YAML Front Matter | | DOCX | .docx | application/vnd.openxmlformats-officedocument.wordprocessingml.document | Word document format, style preservation | | PDF | .pdf | application/pdf | Generated with WeasyPrint, pagination support | | Text | .txt | text/plain | Plain text, basic formatting preserved |

Conversion Matrix

flowchart LR
    subgraph Sources["Source Formats"]
        MD_S[Markdown]
        HTML_S[HTML]
        DOCX_S[DOCX]
        PDF_S[PDF]
        TXT_S[Text]
    end

    subgraph Targets["Target Formats"]
        MD_T[Markdown]
        HTML_T[HTML]
        DOCX_T[DOCX]
        PDF_T[PDF]
        TXT_T[Text]
    end

    MD_S --> Targets
    HTML_S --> Targets
    DOCX_S --> Targets
    PDF_S --> Targets
    TXT_S --> Targets

Installation

Using pip (Recommended)

pip install mcp-document-converter

From Source

git clone https://github.com/xt765/mcp-document-converter.git
cd mcp-document-converter
pip install -e .

MCP Tools

This server provides the following tools:

convert_document

Convert a document from one format to another.

Arguments:

• source_path (string, required): Path to the source document.
• target_format (string, required): Target format (html, pdf, markdown, docx, text).
• output_path (string, optional): Path for the output file.
• source_format (string, optional): Format of the source file (auto-detected if not provided).
• options (object, optional): Additional options like template, css, and preserve_metadata.

Configuration

Using in Trae IDE / Claude Desktop

Add the following to your MCP configuration file:

Option 1: Using PyPI (Recommended)

{
  "mcpServers": {
    "mcp-document-converter": {
      "command": "uvx",
      "args": [
        "mcp-document-converter"
      ]
    }
  }
}

Option 2: Using GitHub repository

{
  "mcpServers": {
    "mcp-document-converter": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/xt765/mcp-document-converter",
        "mcp-document-converter"
      ]
    }
  }
}

Option 3: Using Gitee repository (Faster access in China)

{
  "mcpServers": {
    "mcp-document-converter": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://gitee.com/xt765/mcp-document-converter",
        "mcp-document-converter"
      ]
    }
  }
}

Option 4: Using pip (Manual installation)

First install the package:

pip install mcp-document-converter

Then add to configuration:

{
  "mcpServers": {
    "mcp-document-converter": {
      "command": "mcp-document-converter",
      "args": []
    }
  }
}

Using in Cherry Studio

Cherry Studio is a powerful open-source desktop AI client assistant that supports integrating various tools through the MCP protocol

Configuration Example:

Usage Example:

Usage

As an MCP Tool

After configuration, AI assistants can directly call the following tools:

1. convert_document (Recommended)

Use a unified interface to convert any supported document type.

# Markdown to HTML
convert_document(
    source_path="document.md",
    target_format="html"
)

# HTML to PDF
convert_document(
    source_path="document.html",
    target_format="pdf"
)

# DOCX to Markdown
convert_document(
    source_path="document.docx",
    target_format="markdown"
)

# Conversion with options
convert_document(
    source_path="document.md",
    target_format="html",
    output_path="output.html",
    options={
        "css": "custom.css",
        "preserve_metadata": True
    }
)

2. listsupportedformats

List all supported document formats.

list_supported_formats()

3. getconversionmatrix

Get the complete format conversion matrix.

get_conversion_matrix()

4. can_convert

Check if conversion from source format to target format is supported.

can_convert(source_format="markdown", target_format="pdf")

5. getformatinfo

Get detailed information about a specific format.

get_format_info(format="markdown")

As a Python Library

from mcp_document_converter import DocumentConverter
from mcp_document_converter.registry import get_registry
from mcp_document_converter.parsers import MarkdownParser, HTMLParser
from mcp_document_converter.renderers import HTMLRenderer, PDFRenderer

# Register parsers and renderers
registry = get_registry()
registry.register_parser(MarkdownParser())
registry.register_parser(HTMLParser())
registry.register_renderer(HTMLRenderer())
registry.register_renderer(PDFRenderer())

# Create converter
converter = DocumentConverter(registry)

# Convert document
result = converter.convert(
    source="input.md",
    target_format="html",
    output_path="output.html"
)

if result.success:
    print(f"✅ Conversion successful: {result.output_path}")
else:
    print(f"❌ Conversion failed: {result.error_message}")

Tool Interface Details

convert_document

Convert a document from one format to another.

Parameters:

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | source_path | string | ✅ | Source file path, supports absolute or relative paths | | target_format | string | ✅ | Target format: html, pdf, markdown, docx, text | | output_path | string | ❌ | Output file path (optional, defaults to source filename) | | source_format | string | ❌ | Source format (optional, auto-detected from file extension) | | options | object | ❌ | Conversion options |

Options:

| Option | Type | Default | Description | |--------|------|---------|-------------| | template | string | - | Template name | | css | string | - | Custom CSS styles | | preserve_metadata | boolean | true | Whether to preserve metadata | | extract_images | boolean | true | Whether to extract images |

Example:

{
  "source_path": "/path/to/document.md",
  "target_format": "html",
  "output_path": "/path/to/output.html",
  "options": {
    "css": "body { font-family: Arial; }",
    "preserve_metadata": true
  }
}

Extension Development

Adding a New Parser

from typing import List, Union
from pathlib import Path
from mcp_document_converter.core.parser import BaseParser
from mcp_document_converter.core.ir import DocumentIR, Node, NodeType

class MyParser(BaseParser):
    @property
    def supported_extensions(self) -> List[str]:
        return [".myext"]
    
    @property
    def format_name(self) -> str:
        return "myformat"
    
    @property
    def mime_types(self) -> List[str]:
        return ["application/x-myformat"]
    
    def parse(self, source: Union[str, Path, bytes], **options) -> DocumentIR:
        # Read source file
        content = self._read_source(source)
        
        # Parse into DocumentIR
        document = DocumentIR()
        document.title = "My Document"
        
        # Add content nodes
        document.add_node(Node(
            type=NodeType.PARAGRAPH,
            content=[Node(type=NodeType.TEXT, content="Hello World")]
        ))
        
        return document

Adding a New Renderer

from typing import Any
from mcp_document_converter.core.renderer import BaseRenderer
from mcp_document_converter.core.ir import DocumentIR

class MyRenderer(BaseRenderer):
    @property
    def output_extension(self) -> str:
        return ".myext"
    
    @property
    def format_name(self) -> str:
        return "myformat"
    
    @property
    def mime_type(self) -> str:
        return "application/x-myformat"
    
    def render(self, document: DocumentIR, **options: Any) -> str:
        # Render DocumentIR to target format
        parts = []
        
        if document.title:
            parts.append(f"# {document.title}")
        
        for node in document.content:
            # Render each node
            pass
        
        return "\n".join(parts)

Registering Extensions

from mcp_document_converter.registry import get_registry

# Register new parser and renderer
registry = get_registry()
registry.register_parser(MyParser())
registry.register_renderer(MyRenderer())

Testing

# Run all tests
python tests/test_conversion.py

# Run specific test
python tests/test_conversion.py::test_markdown_to_html

Environment Variables

| Variable | Description | Default | |----------|-------------|---------| | MCP_CONVERTER_LOG_LEVEL | Log level | INFO | | MCP_CONVERTER_TEMP_DIR | Temporary files directory | System temp directory |

Dependencies

Core Dependencies

• mcp >= 1.26.0 - MCP protocol implementation
• pydantic >= 2.12.5 - Data validation

Parser Dependencies

• markdown >= 3.5.0 - Markdown parsing
• beautifulsoup4 >= 4.12.0 - HTML parsing
• python-docx >= 1.1.0 - DOCX parsing
• pypdf >= 6.7.4 - PDF parsing
• chardet >= 5.0.0 - Encoding detection
• pyyaml >= 6.0.0 - YAML parsing

Renderer Dependencies

• weasyprint >= 60.0 - PDF rendering
• pygments >= 2.17.0 - Code highlighting
• jinja2 >= 3.1.6 - Template engine
• reportlab >= 4.0.0 - PDF generation

Development Dependencies

• pytest >= 7.0.0 - Testing framework
• pytest-asyncio >= 0.21.0 - Async testing support
• pytest-cov >= 4.0.0 - Coverage reporting
• basedpyright >= 1.0.0 - Type checking
• ruff >= 0.1.0 - Linting and formatting

License

MIT License

Contributing

Issues and Pull Requests are welcome!

Related Projects

• MCP Document Reader - MCP document reader supporting multiple document formats
• Model Context Protocol - Official Model Context Protocol documentation

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: xt765
• Source: xt765/mcp-document-converter
• License: MIT

Install and usage instructions live in the source repository linked above.