MCP Data Visualization Server

Generate interactive data visualizations from natural language queries on a DuckDB database.

Transform natural language into beautiful, interactive data visualizations using Claude Desktop withDuckDBandDatabrickssupport - one-command installation!

- 🚀One-Command Installation-pip install mcp-visualization-duckdb
- 🔧Automatic Configuration- Automatically sets up Claude Desktop
- 🗣️Natural Language Interface- Chat with Claude to create visualizations
- 📊Interactive Charts- Plotly-powered HTML widgets
- 🏢Databricks Integration- Connect to Databricks SQL warehouses with secure credential management
- 🗃️DuckDB Support- Local database with CSV import and interactive browser
- 🔄Multi-Database- Seamlessly switch between DuckDB and Databricks
- 🧠Rule-Based Analysis- Smart chart suggestions without external LLM dependencies
- 📈Multiple Chart Types- Bar, line, scatter, pie, histogram, box plots, heatmaps, and area charts
- 🔍Smart Insights- Automatic statistical analysis and pattern detection
- 🛡️Security First- SQL injection protection, encrypted credential storage
- ⚡No External Dependencies- Works entirely offline with Claude Desktop

# 1. Install the package (includes DuckDB and all dependencies) pip install mcp-visualization-duckdb # 2. Configure Claude Desktop automatically mcp-viz configure # 3. Restart Claude Desktop and start chatting!

That's it! No manual configuration needed.

Note: DuckDB and all other dependencies are automatically installed with the package. No separate database installation required!

Option B: Databricks (Enterprise Data Warehouse)

# 1. Install the package pip install mcp-visualization-duckdb # 2. Configure Databricks credentials (secure, interactive setup) mcp-viz databricks configure # 3. Configure Claude Desktop (auto-detects Databricks) mcp-viz configure # 4. Restart Claude Desktop and start chatting!

- Databricks workspace hostname (e.g.,your-company.cloud.databricks.com)
- SQL warehouse HTTP path (e.g.,/sql/1.0/warehouses/abc123)
- Personal access token (generated in Databricks User Settings > Developer > Access Tokens)

🔒 Security: Credentials are stored securely using your system keyring or encrypted files. Tokens are never stored in plain text or command history.

- ✅Detects your platform(Windows/Mac/Linux)
- ✅Finds Claude Desktop configautomatically
- ✅Auto-detects database type(DuckDB or Databricks)
- ✅Sets up databasewith sample data (DuckDB)
- ✅Securely stores credentials(Databricks)
- ✅Configures pathscorrectly
- ✅Creates backupsof existing config
- ✅Validates setupto ensure it works

Once installed, you'll see detailed instructions with:

- Next steps for your chosen database type
- Quick start commands
- Sample database links
- Example queries to try

Runmcp-viz-setupanytime to see the post-installation guide again!

# Run these commands ONCE: mcp-viz databricks configure # First time Databricks setup mcp-viz configure # Configure Claude Desktop # Restart Claude Desktop

📋 Daily Usage - No Reconfiguration Needed

# Just open Claude Desktop and start chatting: "What Databricks catalogs are available?" "List tables in the sales catalog" "Create a chart of revenue by region"

mcp-viz databricks status # Shows if credentials exist mcp-viz databricks test # Test connection and browse catalogs mcp-viz status # Shows MCP server configuration

- ❌Token expires(personal access tokens can have expiration dates)
- ❌Change workspaces(different Databricks instance)
- ❌Change SQL warehouses(different HTTP path)
- ❌Credential issues(runmcp-viz databricks removethen reconfigure)

- 🔒Encrypted on diskor in system keyring
- 🔄Automatically loadedwhen Claude Desktop starts
- ⚡No repeated authenticationneeded

📝 Note: After initial setup, you should be able to use Claude Desktop normally without any additional configuration steps!

After installation, simply chat with Claude using natural language:

- "What tables are available?"- List database tables
- "What Databricks catalogs are available?"- Browse Databricks catalogs
- "Show me schemas in the sales catalog"- List schemas
- "Analyze the sales table"- Get table information
- "Show me the top 10 products by revenue"- Query data

- "Create a bar chart of sales by region"- Generate charts
- "Show me the correlation between price and quantity"- Scatter plots
- "Make a pie chart of customer segments"- Category breakdowns
- "Visualize sales trends over time"- Time series analysis

- "Browse databases in my Downloads folder"- Interactive database browser
- "Connect to C:/path/to/mydata.duckdb"- Switch databases
- "Load CSV from Downloads/sales.csv as table 'sales'"- Import data

- "Switch to the marketing catalog"- Navigate catalogs
- "List schemas in the sales catalog"- Browse schemas
- "What database am I connected to?"- Connection info

# Configuration mcp-viz configure # Interactive setup (auto-detects database type) mcp-viz configure --auto # Automatic setup with defaults mcp-viz status # Check configuration status # Management mcp-viz test # Test server functionality mcp-viz remove # Remove server from Claude Desktop # DuckDB Database mcp-viz create-db # Create sample database mcp-viz create-db --path ./my-data.duckdb # Create at specific path # Databricks Integration mcp-viz databricks configure # Setup Databricks credentials (interactive) mcp-viz databricks status # Check Databricks connection status mcp-viz databricks test # Test connection and browse catalogs mcp-viz databricks remove # Remove stored credentials

"Server not found" in Claude Desktop

# Check configuration status mcp-viz status # Reconfigure if needed mcp-viz configure --force # Restart Claude Desktop completely

# Test server functionality mcp-viz test # Check if all dependencies are installed pip install --upgrade mcp-visualization-duckdb

# Create a fresh database with sample data mcp-viz create-db # Check current configuration mcp-viz status

The installer will prompt you for a database location, or you can specify it:

# During configuration mcp-viz configure # Database path [/home/user/.mcp-visualization/data.duckdb]: /path/to/my/data.duckdb

You can easily switch between databases using Claude Desktop:

- "Browse databases in Documents folder"
- "Connect to /path/to/another/database.duckdb"
- "What database am I currently connected to?"

The package automatically creates sample data including:

- Sales Data- 365 days of sales across regions and products
- Customer Data- 1000 customer records with demographics
- Product Data- 100 products with categories and pricing

Want more data to explore? Check out these high-quality sample databases:

- Financial Data- Stock prices, trading data, market analysis
- NYC Taxi- Real NYC taxi trip records with geospatial data
- TPC-H Benchmark- Standard database performance testing data
- E-commerce- Product catalogs, orders, customer behavior
- IoT Sensor Data- Time-series sensor readings and telemetry

Simply download any.duckdbfile and load it:

# Download a sample database, then: "Load database from Downloads/nyc-taxi.duckdb" "What tables are available in this database?" "Create a heatmap of taxi pickups by hour and day"

Perfect for testing, learning, and demonstrating data visualization capabilities!

# Clone repository git clone https://github.com/your-github-username/mcp-visualization-duckdb.git cd mcp-visualization-duckdb # Install in development mode pip install -e . # Configure for development mcp-viz configure # Run tests python test_package.py

SeeDevelopment Guidefor detailed instructions.

mcp-viz databricks status # Shows connection status mcp-viz databricks test # Tests connection and shows catalogs

- PyPI:https://pypi.org/project/mcp-visualization-duckdb/
- Repository:https://github.com/your-github-username/mcp-visualization-duckdb
- License: MIT

- Model Context Protocolby Anthropic
- DuckDBfor fast analytical processing
- Plotlyfor interactive visualizations
- Claude Desktopfor the amazing AI interface

pip install mcp-visualization-duckdb && mcp-viz configure

Maritime intelligence for tracking vessels, analysing ports, and exploring ship data.