Plots MCP Server

A lightweight Model Context Protocol (MCP) server for data visualization. It exposes tools to render charts (line, bar, pie, scatter, heatmap, etc.) from tabular data and returns MCP-compatible image/text content.

Why MCP Plots?

• Instant, visual-first charts using Mermaid (renders directly in MCP clients like Cursor)
• Simple prompts to generate charts from plain data
• Zero-setup options via uvx, or install from PyPI/Docker
• Flexible output formats: mermaid (default), PNG image, or text

Quick Usage

• Ask your MCP client: "Create a bar chart showing sales: A=100, B=150, C=80"
• Default output is Mermaid, so diagrams render instantly in Cursor

Quick Start

PyPI Installation (Recommended)

pip install mcp-plots
mcp-plots  # Start the server

For Cursor Users

1. Install the package: pip install mcp-plots
2. Add to your Cursor MCP config (~/.cursor/mcp.json):

   {
     "mcpServers": {
       "plots": {
         "command": "mcp-plots",
         "args": ["--transport", "stdio"]
       }
     }
   }
   

   Alternative (zero-install via uvx + PyPI):

   {
     "mcpServers": {
       "plots": {
         "command": "uvx",
         "args": ["mcp-plots", "--transport", "stdio"]
       }
     }
   }
   

3. Restart Cursor
4. Ask: "Create a bar chart showing sales: A=100, B=150, C=80"

Development Installation

uvx --from git+https://github.com/mr901/mcp-plots.git run-server.py

Documentation → | Quick Start → | API Reference →

Project layout

src/
  app/                # Server construction and runtime
    server.py
  capabilities/       # MCP tools and prompts
    tools.py
    prompts.py
  visualization/      # Plotting engines and configurations
    chart_config.py
    generator.py

Requirements

• Python 3.10+
• See requirements.txt

Setup Routes

uvx (Recommended)

The easiest way to run the MCP server without managing Python environments:

# Run directly with uvx (no installation needed)
uvx --from git+https://github.com/mr901/mcp-plots.git run-server.py

# Or install and run the command
uvx --from git+https://github.com/mr901/mcp-plots.git mcp-plots

# With custom options
uvx --from git+https://github.com/mr901/mcp-plots.git mcp-plots --port 8080 --log-level DEBUG

Why uvx?

• No Environment Management: Automatically handles Python dependencies
• Isolated Execution: Runs in its own virtual environment
• Always Latest: Pulls fresh code from repository
• Zero Setup: Works immediately without pip install
• Cross-Platform: Same command works on Windows, macOS, Linux

PyPI (Traditional Installation)

1. Install dependencies

pip install -r requirements.txt

2. Run the server (HTTP transport, default port 8000)

python -m src --transport streamable-http --host 0.0.0.0 --port 8000 --log-level INFO

3. Run with stdio (for MCP clients that spawn processes)

python -m src --transport stdio

Local Development (from source)

git clone https://github.com/mr901/mcp-plots.git
cd mcp-plots
pip install -e .
python -m src --transport stdio --log-level DEBUG

Docker

docker build -t mcp-plots .
docker run -p 8000:8000 mcp-plots

Environment variables (optional):

• MCP_TRANSPORT (streamable-http|stdio)
• MCP_HOST (default 0.0.0.0)
• MCP_PORT (default 8000)
• LOG_LEVEL (default INFO)

Tools

• list_chart_types() → returns available chart types
• list_themes() → returns available themes
• suggest_fields(sample_rows) → suggests field roles based on data samples
• render_chart(chart_type, data, field_map, config_overrides?, options?, output_format?) → returns MCP content
• generate_test_image() → generates a test image (red circle) to verify MCP image support

Cursor Integration

This MCP server is fully compatible with Cursor's image support! When you use the render_chart tool:

• Charts appear directly in chat - No need to save files or open separate windows
• AI can analyze your charts - Vision-enabled models can discuss and interpret your visualizations
• Perfect MCP format - Uses the exact base64 PNG format that Cursor expects

The server returns images in the MCP format Cursor requires:

{
  "content": [
    {
      "type": "image", 
      "data": "<base64-encoded-png>",
      "mimeType": "image/png"
    }
  ]
}

Example call (pseudo):

render_chart(
  chart_type="bar",
  data=[{"category":"A","value":10},{"category":"B","value":20}],
  field_map={"category_field":"category","value_field":"value"},
  config_overrides={"title":"Example Bar","width":800,"height":600,"output_format":"MCP_IMAGE"}
)

Return shape (PNG):

{
  "status": "success",
  "content": [{"type":"image","data":"<base64>","mimeType":"image/png"}]
}

Configuration

The server can be configured via environment variables or command line arguments:

Server Settings

• MCP_TRANSPORT - Transport type: streamable-http or stdio (default: streamable-http)
• MCP_HOST - Host address (default: 0.0.0.0)
• MCP_PORT - Port number (default: 8000)
• LOG_LEVEL - Logging level: DEBUG, INFO, WARNING, ERROR, CRITICAL (default: INFO)
• MCP_DEBUG - Enable debug mode: true or false (default: false)

Chart Settings

• CHART_DEFAULT_WIDTH - Default chart width in pixels (default: 800)
• CHART_DEFAULT_HEIGHT - Default chart height in pixels (default: 600)
• CHART_DEFAULT_DPI - Default chart DPI (default: 100)
• CHART_MAX_DATA_POINTS - Maximum data points per chart (default: 10000)

Command Line Usage

With uvx (recommended):

uvx --from git+https://github.com/mr901/mcp-plots.git mcp-plots --help

# Examples:
uvx --from git+https://github.com/mr901/mcp-plots.git mcp-plots --port 8080 --log-level DEBUG
uvx --from git+https://github.com/mr901/mcp-plots.git mcp-plots --chart-width 1200 --chart-height 800

Traditional Python:

python -m src --help

# Examples:
python -m src --transport streamable-http --host 0.0.0.0 --port 8000
python -m src --log-level DEBUG --chart-width 1200 --chart-height 800

Docker

Build image:

docker build -t mcp-plots .

Run container with custom configuration:

docker run --rm -p 8000:8000 \
  -e MCP_TRANSPORT=streamable-http \
  -e MCP_HOST=0.0.0.0 \
  -e MCP_PORT=8000 \
  -e LOG_LEVEL=INFO \
  -e CHART_DEFAULT_WIDTH=1000 \
  -e CHART_DEFAULT_HEIGHT=700 \
  -e CHART_DEFAULT_DPI=150 \
  -e CHART_MAX_DATA_POINTS=5000 \
  mcp-plots

Cursor MCP Integration

Quick Setup for Cursor

The Plots MCP Server is designed to work seamlessly with Cursor's MCP support. Here's how to integrate it:

1. Add to Cursor's MCP Configuration

Add this to your Cursor MCP configuration file (~/.cursor/mcp.json or similar):

{
  "mcpServers": {
    "plots": {
      "command": "uvx",
      "args": [
        "--from", 
        "git+https://github.com/mr901/mcp-plots.git@main",
        "mcp-plots",
        "--transport", 
        "stdio"
      ],
      "env": {
        "LOG_LEVEL": "INFO",
        "CHART_DEFAULT_WIDTH": "800",
        "CHART_DEFAULT_HEIGHT": "600"
      }
    }
  }
}

2. Alternative: HTTP Transport

For HTTP-based integration:

{
  "mcpServers": {
    "plots-http": {
      "command": "uvx",
      "args": [
        "--from", 
        "git+https://github.com/mr901/mcp-plots.git@main", 
        "mcp-plots",
        "--transport", 
        "streamable-http",
        "--host", 
        "127.0.0.1",
        "--port", 
        "8000"
      ]
    }
  }
}

3. Local Development Setup

For local development (if you have the code cloned):

{
  "mcpServers": {
    "plots-dev": {
      "command": "python",
      "args": ["-m", "src", "--transport", "stdio"],
      "cwd": "/path/to/mcp-plots",
      "env": {
        "LOG_LEVEL": "DEBUG"
      }
    }
  }
}

4. Verify Integration

After adding the configuration:

1. Restart Cursor
2. Check MCP connection in Cursor's MCP panel
3. Test with a simple chart:

   Create a bar chart showing sales data: A=100, B=150, C=80
   

MERMAID-First Approach

This server prioritizes MERMAID output by default because:

• ✅ Renders instantly in Cursor - No external viewers needed
• ✅ Interactive - Cursor can analyze and discuss the diagrams
• ✅ Lightweight - Fast generation and display
• ✅ Scalable - Vector-based, works at any zoom level

Chart Types with Native MERMAID Support:

• line, bar, pie, area → xychart-beta format
• histogram → xychart-beta with automatic binning
• funnel → Styled flowchart with color gradients
• gauge → Flowchart with color-coded value indicators
• sankey → Flow diagrams with source/target styling

Available Tools

render_chart

Main chart generation tool with MERMAID-first approach.

Parameters:

• chart_type - Chart type (line, bar, pie, scatter, heatmap, etc.)
• data - List of data objects
• field_map - Field mappings (x_field, y_field, category_field, etc.)
• config_overrides - Chart configuration overrides
• output_format - Output format (mermaid [default], mcp_image, mcp_text)

Special Modes:

• chart_type="help" - Show available chart types and themes
• chart_type="suggest" - Analyze data and suggest field mappings

configure_preferences

Interactive configuration tool for setting user preferences.

Parameters:

• output_format - Default output format (mermaid, mcp_image, mcp_text)
• theme - Default theme (default, dark, seaborn, minimal)
• chart_width - Default chart width in pixels
• chart_height - Default chart height in pixels
• reset_to_defaults - Reset all preferences to system defaults

Features:

• Persistent Settings - Saved to ~/.plots_mcp_config.json
• Live Preview - Shows sample chart with current settings
• Override Support - Use config_overrides for one-off changes

Documentation

Additional Resources

• Complete Documentation - Technical documentation hub
• Quick Start - 5-minute setup guide
• Integration Guide - MCP client setup and configuration
• API Reference - Complete tool specifications and examples
• Advanced Guide - Architecture, deployment, and development
• Sample Prompts - Ready-to-use testing examples

Chart Examples

Basic Bar Chart:

{
  "chart_type": "bar",
  "data": [
    {"category": "Sales", "value": 120},
    {"category": "Marketing", "value": 80},
    {"category": "Support", "value": 60}
  ],
  "field_map": {
    "category_field": "category", 
    "value_field": "value"
  }
}

Time Series Line Chart:

{
  "chart_type": "line",
  "data": [
    {"date": "2024-01", "revenue": 1000},
    {"date": "2024-02", "revenue": 1200},
    {"date": "2024-03", "revenue": 1100}
  ],
  "field_map": {
    "x_field": "date",
    "y_field": "revenue"
  }
}

Funnel Chart:

{
  "chart_type": "funnel",
  "data": [
    {"stage": "Awareness", "value": 1000},
    {"stage": "Interest", "value": 500}, 
    {"stage": "Purchase", "value": 100}
  ],
  "field_map": {
    "category_field": "stage",
    "value_field": "value"
  }
}

🔧 Configuration

Environment Variables

• MCP_TRANSPORT - Transport type (streamable-http | stdio)
• MCP_HOST - Host address (default: 0.0.0.0)
• MCP_PORT - Port number (default: 8000)
• LOG_LEVEL - Logging level (default: INFO)
• MCP_DEBUG - Enable debug mode (true | false)
• CHART_DEFAULT_WIDTH - Default chart width in pixels (default: 800)
• CHART_DEFAULT_HEIGHT - Default chart height in pixels (default: 600)
• CHART_DEFAULT_DPI - Default chart DPI (default: 100)
• CHART_MAX_DATA_POINTS - Maximum data points per chart (default: 10000)

User Preferences

Personal preferences are stored in ~/.plots_mcp_config.json:

{
  "defaults": {
    "output_format": "mermaid",
    "theme": "default",
    "chart_width": 800,
    "chart_height": 600
  },
  "user_preferences": {
    "output_format": "mcp_image",
    "theme": "dark"
  }
}

🚀 Advanced Usage

Custom Themes

Available themes: default, dark, seaborn, minimal, whitegrid, darkgrid, ticks

High-Resolution Charts

uvx --from git+https://github.com/mr901/mcp-plots.git mcp-plots \
  --chart-width 1920 \
  --chart-height 1080 \
  --chart-dpi 300

Performance Optimization

• Use max_data_points to limit large datasets
• MERMAID output is fastest for quick visualization
• PNG output for high-quality static images
• SVG output for scalable vector graphics

🐛 Troubleshooting

Common Issues

Issue: Charts not rendering in Cursor

• Solution: Ensure output_format="mermaid" (default)
• Check: MCP server connection in Cursor

Issue: uvx command not found

• Solution: Install uv: curl -LsSf https://astral.sh/uv/install.sh | sh

Issue: Port already in use

• Solution: Use different port: --port 8001

Issue: Large datasets slow

• Solution: Sample data or increase --max-data-points

Debug Mode

uvx --from git+https://github.com/mr901/mcp-plots.git mcp-plots \
  --debug \
  --log-level DEBUG

📝 Notes

• Matplotlib runs headless (Agg backend) in the container
• For large datasets, sample your data for responsiveness
• Chart defaults can be overridden per-request via config_overrides
• MERMAID charts render instantly in Cursor for the best user experience
• User preferences persist across sessions and apply to all charts by default