DocuMCP
A document management system with MCP integration, enabling Claude Desktop to upload, search, and access documents via natural language.
README
DocuMCP - Document Management System with MCP Integration
A powerful document management system that combines local file storage with vector database capabilities, accessible through both a web GUI and Claude Desktop via the Model Context Protocol (MCP).
Features
- 📚 Document Management: Upload, store, and organize documents with a beautiful web interface
- 🔍 Vector Search: Semantic search powered by Qdrant vector database
- 🤖 MCP Integration: Access documents directly from Claude Desktop
- 🔐 Authentication: Secure JWT-based authentication system
- 🎨 Modern GUI: Responsive web dashboard for easy document management
- 🐳 Docker Support: Easy deployment with Docker Compose
Quick Start
Prerequisites
- Node.js 20+
- Docker & Docker Compose (optional, for Qdrant)
- Claude Desktop (for MCP integration)
Installation
- Clone and install dependencies:
git clone <repository>
cd documcp
npm install
- Start Qdrant (using Docker):
docker run -p 6333:6333 -v ./qdrant_data:/qdrant/storage qdrant/qdrant
- Configure environment:
cp .env.example .env
# Edit .env with your settings
- Build and start the application:
npm run build
npm start
- Access the web interface: Open http://localhost:3456 in your browser
- Default login:
admin/admin123
Docker Deployment
For production deployment with Docker:
docker-compose up -d
This will start:
- Qdrant vector database on port 6333
- DocuMCP web server on port 3456
- MCP server on port 3457
Claude Desktop Integration
- Build the MCP server:
npm run build
- Add to Claude Desktop configuration:
Edit your Claude Desktop config file:
- Mac:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Add this configuration:
{
"mcpServers": {
"documcp": {
"command": "node",
"args": ["/path/to/documcp/dist/mcp-server.js"],
"env": {
"QDRANT_URL": "http://localhost:6333",
"COLLECTION_NAME": "documents"
}
}
}
}
- Restart Claude Desktop
Available MCP Tools
Once configured, Claude can use these tools:
list_documents- List all documents in the systemsearch_documents- Semantic search across documentsread_document- Read specific document contentdelete_document- Remove documents from the system
Architecture
DocuMCP/
├── Web GUI (Express + HTML/JS)
│ ├── Authentication (JWT)
│ ├── File Upload/Download
│ └── Document Management
├── MCP Server
│ └── Document Access Tools
├── Storage Layer
│ ├── Local File System (./uploads)
│ └── Qdrant Vector DB
└── API Layer
├── REST API
└── MCP Protocol
Configuration
Environment Variables
| Variable | Description | Default |
|---|---|---|
PORT |
Web server port | 3456 |
JWT_SECRET |
JWT signing secret | (required) |
ADMIN_PASSWORD |
Admin login password | admin123 |
QDRANT_URL |
Qdrant server URL | http://localhost:6333 |
COLLECTION_NAME |
Qdrant collection | documents |
UPLOAD_DIR |
File storage directory | ./uploads |
MAX_FILE_SIZE |
Max upload size (bytes) | 10485760 |
Development
# Run in development mode
npm run dev
# Run MCP server standalone
npm run mcp:dev
# Build TypeScript
npm run build
Security Notes
⚠️ Important for Production:
- Change default
JWT_SECRETandADMIN_PASSWORD - Use HTTPS in production
- Configure proper CORS settings
- Implement rate limiting
- Add file type restrictions
- Use environment-specific configs
Troubleshooting
Qdrant Connection Issues
- Ensure Qdrant is running:
docker ps - Check logs:
docker logs qdrant - Verify port 6333 is accessible
MCP Server Not Connecting
- Check Claude Desktop logs
- Verify path in config is absolute
- Ensure Node.js is in PATH
Upload Failures
- Check
MAX_FILE_SIZEsetting - Verify
uploadsdirectory permissions - Check available disk space
License
MIT
Contributing
Pull requests welcome! Please ensure:
- Code follows TypeScript best practices
- Tests pass (when implemented)
- Documentation is updated
- Security considerations addressed
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。