SQL数据分析服务器

You: "Load sales_2024.csv as sales and show me the top 10 products by revenue"

Claude/ChatGPT:
1. Imports your CSV into a DuckDB database
2. Writes SQL: SELECT product, SUM(revenue) FROM sales GROUP BY product ORDER BY revenue DESC LIMIT 10
3. Executes the query
4. Shows results with insights: "Your top product is Widget A with $50K revenue..."

# Clone or download this repo
cd mcp-data-analyst

# Start the container
./docker-start.sh

# Falls das File nicht ausführbar ist. führe im Terminal folgenden Befehl aus:

chmod +x docker-start.sh

# Danach starte es erneut:

./docker-start.sh 


# Follow the instructions to configure Claude/ChatGPT

# Clone or download this repo
cd mcp-data-analyst

# Install dependencies
pip install -r requirements.txt

# Run the server
python server.py

{
  "mcpServers": {
    "mcp-data-analyst": {
      "command": "docker",
      "args": ["exec", "-i", "data-analyst-mcp", "python3", "server.py"]
    }
  }
}

{
  "mcpServers": {
    "mcp-data-analyst": {
      "command": "python3",
      "args": ["/absolute/path/to/mcp-data-analyst/server.py"]
    }
  }
}

# Copy the example
cp .env.example .env

# Edit .env and add your token
nano .env

MOTHERDUCK_TOKEN=your_token_here
MEMORY_LIMIT=4G
CPU_LIMIT=2.0

You: "Load example_data.csv as customers and show me the data"

You: "Load my sales data from https://example.com/sales.csv as sales"
AI: ✓ Imported 50,000 rows into 'sales' table

You: "What are the top 5 products by revenue?"
AI: [Writes and executes SQL, shows results with insights]

You: "Show me monthly revenue trends"
AI: [Creates time-series analysis with DATE_TRUNC]

You: "Load sales.csv as sales and products.csv as products"
AI: ✓ Imported both tables

You: "Join these tables and show me which product categories generate the most revenue"
AI: [Automatically identifies the relationship (product_id), performs JOIN, provides analysis]

You: "Analyze customer behavior and create segments based on purchase patterns"
AI: [Performs multi-step analysis with CTEs, window functions, and provides business insights]

You: "Load this Google Sheet: https://docs.google.com/spreadsheets/d/abc123/edit"
AI: ✓ Converted to CSV and imported

You: "Calculate the correlation between marketing spend and sales"
AI: [Performs statistical analysis]

# Start server
./docker-start.sh

# Stop server
./docker-stop.sh

# View logs
docker-compose logs -f

# Restart
docker-compose restart

# Stop and remove
docker-compose down

# Rebuild
docker-compose down && docker-compose up -d --build

chmod +x docker-start.sh docker-stop.sh setup.sh

memory: 8G  # or 16G
cpus: '4.0'

┌─────────────────────────────────────┐
│   Claude Desktop / ChatGPT Desktop  │
│   (Natural Language Interface)      │
└─────────────────┬───────────────────┘
                  │ MCP Protocol
                  ▼
┌─────────────────────────────────────┐
│   MCP Data Analyst Server (Python) │
│   - 6 Tools (FastMCP)               │
│   - CSV Import & SQL Query          │
└─────────────────┬───────────────────┘
                  │
                  ▼
┌─────────────────────────────────────┐
│   DuckDB Database                   │
│   - In-Memory (< 10M rows)          │
│   - MotherDuck Cloud (billions)     │
└─────────────────────────────────────┘

deploy:
  resources:
    limits:
      memory: 8G      # Increase for large datasets
      cpus: '4.0'     # More CPU for faster processing

You: "Load sales.csv, products.csv, and customers.csv"
AI: [Imports all three as separate tables]

You: "Show me how these tables relate to each other"
AI: [Analyzes schemas, identifies foreign keys, suggests JOINs]

You: "Export the top 100 customers to a CSV file"
AI: [Executes query and saves to /data/top_customers.csv]

@mcp.tool()
def my_custom_tool(param: str, count: int = 10) -> dict:
    """Description that appears to the AI"""
    # Your logic here
    return {"result": "..."}

mcp-data-analyst/
├── server.py              # Main MCP server (FastMCP)
├── requirements.txt       # Python dependencies
├── Dockerfile            # Container definition
├── docker-compose.yml    # Docker orchestration
├── docker-start.sh       # Easy start script
├── docker-stop.sh        # Easy stop script
├── setup.sh              # Direct install script
├── .env.example          # Environment template
├── .gitignore            # Git ignore rules
└── example_data.csv      # Sample data for testing