DocNav MCP Server

DocNav is a Model Context Protocol (MCP) server which empowers LLM Agents to read, analyze, and manage lengthy documents intelligently, mimicking human-like comprehension and navigation capabilities.

Features

• Document Navigation: Navigate through document sections, headings, and content structure
• Content Extraction: Extract and summarize specific document sections
• Search & Query: Find specific content within documents using intelligent search
• Multi-format Support: Currently supports Markdown (.md) files, with planned support for PDF and other formats
• MCP Integration: Seamless integration with MCP-compatible LLMs and applications

Architecture

DocNav follows a modular, extensible architecture:

• Core MCP Server: Main server implementation using the MCP protocol
• Document Processors: Pluggable processors for different file types
• Navigation Engine: Handles document structure analysis and navigation
• Content Extractors: Extract and format content from documents
• Search Engine: Provides search and query capabilities across documents

Installation

Prerequisites

• Python 3.10+
• uv package manager

Setup

1. Clone the repository:

git clone https://github.com/shenyimings/DocNav-MCP.git
cd DocNav-MCP

2. Install dependencies:

uv sync

Usage

Starting the MCP Server

uv run server.py

Connect to the MCP server

{
  "mcpServers": {
    "docnav": {
      "command": "{{PATH_TO_UV}}", // Run `which uv` and place the output here
      "args": [
        "--directory",
        "{{PATH_TO_SRC}}",
        "run",
        "server.py"
      ]
    }
  }
}

Available Tools

• load_document: Load a document for navigation and analysis

  • Args: file_path (path to document file)
  • Returns: Success message with auto-generated document ID

• get_outline: Get document outline/table of contents

  • Args: doc_id (document identifier), max_depth (max heading depth, default 3)
  • Returns: Formatted document outline
  • Tip: Use first after loading a document to understand structure

• read_section: Read content of a specific document section

  • Args: doc_id (document identifier), section_id (e.g., 'h1_0', 'h2_1')
  • Returns: Section content with subsections

• search_document: Search for specific content within a document

  • Args: doc_id (document identifier), query (search term or phrase)
  • Returns: Formatted search results with context

• navigate_section: Get navigation context for a section

  • Args: doc_id (document identifier), section_id (section to navigate to)
  • Returns: Navigation context with parent, siblings, children

• list_documents: List all currently loaded documents

  • Returns: List of loaded documents with metadata

• get_document_stats: Get statistics about a loaded document

  • Args: doc_id (document identifier)
  • Returns: Document statistics and structure info

• remove_document: Remove a document from the navigator

  • Args: doc_id (document identifier)
  • Returns: Success or error message

Example Usage

# Load a document
result = await tools.load_document("path/to/document.md")

# Get document outline
outline = await tools.get_outline(doc_id)

# Get specific section content
section = await tools.read_section(doc_id, section_id)

# Search within document
results = await tools.search_document(doc_id, "search query")

Development

Project Structure

docnav-mcp/
--- server.py             # Main MCP server
--- docnav/
------- __init__.py           # Package initialization
------- models.py             # Data models
------- navigator.py          # Document navigation engine
------- processors/
------- __init__.py       # Processor package
------- base.py           # Base processor interface
------- markdown.py       # Markdown processor
--- tests/
------- ...                   # Test files

Development Guidelines

See CLAUDE.md for detailed development guidelines including:

• Code quality standards
• Testing requirements
• Package management with uv
• Formatting and linting rules

Adding New Document Processors

1. Create a new processor class inheriting from BaseProcessor
2. Implement the required methods: can_process, process, extract_section, search
3. Register the processor in the DocumentNavigator
4. Add comprehensive tests

Running Tests

# Run all tests
uv run tests/run_tests.py

Code Quality

# Format code
uv run --frozen ruff format .

# Check linting
uv run --frozen ruff check .

# Type checking
uv run --frozen pyright

Roadmap

• [x] Complete Markdown processor implementation
• [x] Add PDF document support (PyMuPDF)
• [x] Improve test coverage and quality
• [ ] Implement advanced search capabilities
• [ ] Add document summarization features
• [ ] Support for additional document formats (DOCX, TXT, etc.)
• [ ] Performance optimizations for large documents
• [ ] Caching mechanisms for frequently accessed documents
• [ ] Add persistent storage for loaded documents

Contributing

1. Fork the repository
2. Create a feature branch
3. Follow the development guidelines in CLAUDE.md
4. Add tests for new functionality
5. Submit a pull request

License

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

Support

For issues and questions:

• Open an issue on GitHub
• Check the documentation in CLAUDE.md
• Review existing issues and discussions