What is Docu Mcp

DocuMCP is an MCP server designed for intelligent code documentation generation, leveraging RAG (Retrieval-Augmented Generation) capabilities to enhance documentation processes.

Use cases

Use cases for DocuMCP include generating user manuals, creating technical documentation, providing code explanations, and producing architectural diagrams for software projects.

How to use

To use DocuMCP, clone the repository, install the necessary packages, build the MCP server, configure Claude Desktop with the appropriate settings, and start any required services.

Key features

Key features include automatic documentation generation, semantic search across code and documentation, architectural diagram creation, user guide generation, support for multiple vector databases, and flexible embedding providers.

Where to use

DocuMCP can be utilized in software development environments, particularly for projects that require comprehensive documentation and efficient code management.

DocuMCP

🤖 A comprehensive MCP system for intelligent code documentation generation with RAG capabilities and multi-agent orchestration

DocuMCP consists of two complementary MCP servers:

1. DocuMCP Server: Core documentation generation with vector embeddings and semantic search
2. DocuMCP Manager: Agent orchestration for parallel documentation workflows using multiple Claude Code sub-agents

Together, they enable Claude to generate, search, and manage documentation for your codebase at any scale, from single files to entire enterprise applications.

✨ Features

Core Documentation Features

• 📚 Generate and update documentation based on your codebase
• 🔍 Semantic search across code, documentation, and diagrams
• 📊 Create and merge architectural diagrams
• 📝 Generate user guides
• 💾 Support for multiple vector databases (LanceDB, ChromaDB, Qdrant)
• 🧠 Flexible embedding providers (built-in or Ollama)

Multi-Agent Orchestration (Manager Server)

• 🤖 Spawn multiple Claude Code sub-agents for parallel processing
• 📊 Monitor agent status and retrieve results
• 🔄 Shared vector database across all agents
• ⚡ Scale documentation generation for large codebases
• 💰 Track costs and performance metrics

🚀 Quick Start

Installation via NPX (Recommended)

The easiest way to use DocuMCP is to configure Claude Desktop with the published npm package:

Add the following to your Claude Desktop configuration:

• MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%/Claude/claude_desktop_config.json

For Core DocuMCP Server:

{
  "mcpServers": {
    "docu-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@myjungle/docu-mcp-server"
      ]
    }
  }
}

For DocuMCP Manager (Agent Orchestration):

{
  "mcpServers": {
    "docu-mcp-manager": {
      "command": "npx",
      "args": [
        "-y",
        "@myjungle/docu-mcp-manager"
      ]
    }
  }
}

Restart Claude Desktop and both servers will be available.

Alternative Installation Methods

Using Smithery CLI

Install the server via Smithery CLI:

# Install Smithery CLI if you don't have it
npm install -g @smithery/cli

# Then install the Docu MCP server
npx -y @smithery/cli@latest install @YannickTM/docu-mcp --client claude

🚀 Manual Start

1. Clone and Install

git clone https://github.com/YannickTM/docu-mcp
cd docu-mcp
npm install

2. Build the Servers

# Build DocuMCP Server
cd mcp
npm run build
cd ..

# Build DocuMCP Manager
cd manager
npm run build
cd ..

3. Advanced Configuration

Add the following to your Claude Desktop configuration:

• MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%/Claude/claude_desktop_config.json

Configuration for Both Servers:

{
  "mcpServers": {
    "docu-mcp": {
      "command": "node",
      "env": {
        "VECTOR_DB_PROVIDER": "qdrant",
        "QDRANT_URL": "http://localhost:6333",
        "EMBEDDING_PROVIDER": "ollama",
        "EMBEDDING_MODEL": "bge-m3:latest",
        "EMBEDDING_DIMENSION": "1024",
        "OLLAMA_URL": "http://localhost:11434"
      },
      "args": [
        "/absolute/path/to/DocuMCP/mcp/dist/index.js"
      ]
    },
    "docu-mcp-manager": {
      "command": "node",
      "env": {
        "VECTOR_DB_PROVIDER": "qdrant",
        "QDRANT_URL": "http://localhost:6333",
        "EMBEDDING_PROVIDER": "ollama",
        "EMBEDDING_MODEL": "bge-m3:latest",
        "EMBEDDING_DIMENSION": "1024",
        "OLLAMA_URL": "http://localhost:11434",
        "SUB_AGENT_MODEL": "claude-3-7-sonnet-latest"
      },
      "args": [
        "/absolute/path/to/DocuMCP/manager/dist/index.js"
      ]
    }
  }
}

Important: Both servers should use the same vector database configuration to enable shared access.

4. Start Required Services (if using external providers)

For Qdrant:

cd qdrant
npm run start

For ChromaDB:

cd chromadb
npm run start

5. Restart Claude Desktop

Restart Claude Desktop to load the new configuration.

🛠️ Configuration Options

Vector Database Providers

Embedding Providers

🔧 Available Tools

DocuMCP Server Tools

• 📁 File Operations: read_file, write_file, create_directory, read_directory
• 🔎 Search Tools: search_codebase, search_documentation, search_diagram, search_user_guide
• 📚 Documentation: generate_documentation, generate_user_guide, explain_code
• 📊 Diagrams: generate_diagram, merge_diagram
• 🗃️ Indexing: index_file, index_directory
• 🔀 Merging: merge_documentation

DocuMCP Manager Tools (includes all above plus):

• 🤖 Agent Orchestration:
  • spawn_agent: Create Claude Code sub-agents for documentation tasks
  • manage_agent: Monitor, control, and retrieve results from agents

📋 Requirements

• Node.js 20.11.24+
• Claude Desktop
• (Optional) Docker for running external vector databases

🤝 Contributing

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

Made with ❤️