Documentation Index MCP

Universal MCP (Model Context Protocol) server for accessing and searching library documentation. Enables LLM agents to retrieve and analyze documentation for any configured library using advanced text ranking algorithms.

Features

• Multi-Library Support: Access documentation for multiple configured libraries simultaneously
• Advanced Search: Keyword-based document retrieval with BM25 ranking algorithm for precise results
• Intelligent Document Processing: Parse and process markdown documentation with token estimation
• Semantic Chunking: Split documents into semantically meaningful chunks for better context management
• Token Management: Efficient token counting and management for LLM context windows
• Metadata Extraction: Automatic extraction of document structure and metadata (headings, code blocks, tables)
• Category-Based Weighting: Intelligent ranking based on document categories and sections
• Flexible Configuration: Support for multiple search modes and customizable keyword weighting
• Built with TypeScript: Full type safety and comprehensive test coverage

Installation

NPM

npm install -g package7-mcp

Smithery

To install Documentation Index MCP Server for any client automatically via Smithery:

npx -y @smithery/cli@latest install package7-mcp --client <CLIENT_NAME>

Available clients: cursor, claude, vscode, windsurf, cline, zed, etc.

Example for Cursor:

npx -y @smithery/cli@latest install package7-mcp --client cursor

From Source

# Install dependencies
pnpm install

# Build the project
pnpm build

# Run the server
pnpm start

MCP Client Integration

Documentation Index MCP can be integrated with various AI coding assistants and IDEs that support the Model Context Protocol (MCP).

Requirements

• Node.js >= v18.0.0
• pnpm >= v8.0.0
• An MCP-compatible client (Cursor, Claude Code, VS Code, Windsurf, etc.)

Go to: Settings -> Cursor Settings -> MCP -> Add new global MCP server

Add the following configuration to your ~/.cursor/mcp.json file:

{
  "mcpServers": {
    "docs-index": {
      "command": "npx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

Run this command:

claude mcp add docs-index -- npx -y package7-mcp

Add this to your VS Code MCP config file. See VS Code MCP docs for more info.

"mcp": {
  "servers": {
    "docs-index": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

Add this to your Windsurf MCP config file:

{
  "mcpServers": {
    "docs-index": {
      "command": "npx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

1. Open Cline
2. Click the hamburger menu icon (☰) to enter the MCP Servers section
3. Choose Remote Servers tab
4. Click the Edit Configuration button
5. Add docs-index to mcpServers:

{
  "mcpServers": {
    "docs-index": {
      "command": "npx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

Open Claude Desktop developer settings and edit your claude_desktop_config.json file:

{
  "mcpServers": {
    "docs-index": {
      "command": "npx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

Add this to your Zed settings.json:

{
  "context_servers": {
    "docs-index": {
      "source": "custom",
      "command": "npx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

Add this to your Roo Code MCP configuration file:

{
  "mcpServers": {
    "docs-index": {
      "command": "npx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

{
  "mcpServers": {
    "docs-index": {
      "command": "bunx",
      "args": ["-y", "package7-mcp"]
    }
  }
}

Usage

The MCP server provides the following tools:

get-library-list

Retrieve the list of available libraries and their metadata.

Parameters: None

Returns:

{
  "libraries": [
    {
      "id": "react",
      "name": "React",
      "version": "18.0.0",
      "description": "A JavaScript library for building user interfaces"
    }
  ]
}

get-documents

Search library documentation with keyword-based retrieval using BM25 ranking algorithm.

Parameters:

{
  "libraryId": "react",
  "query": "hooks state management",
  "limit": 10
}

Returns: Array of documents with scores and metadata:

{
  "documents": [
    {
      "id": "doc-123",
      "title": "Using the State Hook",
      "content": "...",
      "score": 0.95,
      "metadata": {
        "section": "Hooks",
        "category": "Advanced"
      }
    }
  ]
}

document-by-id

Get full document content by ID.

Parameters:

{
  "documentId": "doc-123"
}

Returns:

{
  "id": "doc-123",
  "title": "Using the State Hook",
  "content": "Complete document content...",
  "metadata": {
    "tokens": 1250,
    "section": "Hooks"
  }
}

Usage Examples

Example 1: Search for documentation

In Cursor/Claude Code:

Search for React documentation about hooks and find examples of useState usage

Example 2: Get specific library information

In any MCP client:

What libraries are available and show me the latest React documentation

Example 3: Analyze code patterns

In Cursor/Claude Code:

I need to understand how to manage state in React.
Search the React documentation for state management patterns.

Architecture

The project follows a modular architecture:

• constants/: Configuration constants and base prompts
• document/: Document processing logic
  • splitter/: Markdown parsing and document chunking utilities
  • parser/: Specialized parsers for different node types
• repository/: Data access layer and document repositories
• schema/: Zod schemas for runtime type validation
• tool/: MCP tool implementations
• server.ts: Main server entry point

Core Components

• MarkdownSplitter: Intelligent markdown document splitting with semantic awareness
• BM25Calculator: Advanced ranking algorithm for search relevance
• TokenEstimator: Efficient token counting for context management
• DocumentLoader: Flexible document loading and caching
• ChunkConverter: Conversion between different document formats

Development

Setup

# Install dependencies
pnpm install

# Run in development mode (watch mode)
pnpm dev

# Type check
pnpm typecheck

Building

pnpm build

Code Quality

# Lint code
pnpm lint

# Fix linting issues
pnpm lint:fix

# Format code with Prettier
pnpm format

Testing

# Run all tests
pnpm test

# Run tests in watch mode
pnpm test:watch

# Generate coverage report
pnpm test:coverage

The project includes comprehensive test coverage for:

• Document processing and parsing
• BM25 ranking algorithm
• Token estimation
• Search functionality
• Repository operations

Project Structure

src/
├── constants/              # Configuration constants and prompts
│   ├── base-prompt.ts      # Base system prompts
│   ├── category.ts         # Category definitions
│   ├── keyword-weight-config.ts # Search weight configuration
│   └── search-mode.ts      # Search mode definitions
├── document/               # Document processing logic
│   ├── splitter/           # Markdown splitting utilities
│   │   ├── markdown-splitter.ts  # Main splitter
│   │   ├── parser/         # Node-type specific parsers
│   │   └── extractMetadata.ts    # Metadata extraction
│   ├── token-estimator.ts  # Token counting
│   ├── document-loader.ts  # Document loading
│   ├── chunk-converter.ts  # Format conversion
│   └── __test__/           # Document tests
├── repository/             # Data access layer
│   ├── docs.repository.ts  # Document repository
│   └── createDocsRepository.ts # Factory function
├── schema/                 # Zod schemas
│   └── get-document-schema.ts   # Request schemas
├── tool/                   # MCP tool implementations
│   └── tools.ts            # Tool definitions
└── server.ts               # Main server entry point

Configuration

Search Modes

The server supports multiple search modes configured via constants:

• BM25: Advanced relevance ranking (default)
• Keyword: Simple keyword matching
• Semantic: Context-aware search

Category Weighting

Documents can be weighted by category for better ranking:

const categoryWeights = {
  'Getting Started': 1.2,
  'API': 1.0,
  'Examples': 0.9
};

Scripts

• pnpm build - Build the project with TypeScript
• pnpm dev - Development mode with watch
• pnpm start - Start the MCP server
• pnpm test - Run all tests
• pnpm test:watch - Run tests in watch mode
• pnpm test:coverage - Generate coverage report
• pnpm lint - Lint code with ESLint
• pnpm lint:fix - Fix linting issues
• pnpm format - Format code with Prettier
• pnpm typecheck - Type check without building

Docker Support

Build Docker Image

docker build -t package7-mcp .

Run with Docker

docker run -d -p 3000:3000 \
  --name docs-index \
  package7-mcp

Docker Compose Example

Create a docker-compose.yml:

version: '3.8'

services:
  docs-index:
    build: .
    ports:
      - "3000:3000"
    environment:
      - NODE_ENV=production
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "node", "-e", "require('http').get('http://localhost:3000/mcp', (r) => {process.exit(r.statusCode === 200 ? 0 : 1)})"]
      interval: 30s
      timeout: 3s
      retries: 3
      start_period: 5s

Run with Docker Compose:

docker-compose up -d

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT

Author

choesumin