Data Visualization

This MCP server provides an interface for visualizing data using Vega-Lite syntax. It exposes two core tools that let you save data tables and render Vega-Lite visualizations, making it easy to explore and share data visuals from your LLM-driven workflows.

How to use

You interact with the server through an MCP client to save data tables and visualize them with Vega-Lite. First save a data table under a descriptive name, then request a Vega-Lite visualization by supplying the table name and a Vega-Lite spec. The server can return either a text artifact that includes the full Vega-Lite specification or a PNG image representation of the visualization, depending on how you configure the client.

Typical workflow patterns: save a data table, then render multiple visualizations from the same data with different Vega-Lite specs. You can switch the output type to text for easy inspection of the generated spec, or to png for a ready-to-share image.

Example usage with your MCP client: you configure the client to run the datapath and specify the output type you want. The provided example shows how to integrate the server into an MCP environment that can produce PNG or text results.

Example client configuration

{
  "mcpServers": {
    "datavis": {
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/mcp-datavis-server",
        "run",
        "mcp_server_datavis",
        "--output_type",
        "png" 	# or "text"
      ]
    }
  }
}

What you can save and visualize

Save a data table that represents your aggregations under a name you choose. The data should be an array of objects where each object represents a row in the table.

Visualize by providing the data table name and a Vega-Lite specification in JSON format. The server returns either a text artifact with the full spec and data, or a PNG image depending on the configured output type.

Available tools

save_data

Save a table of data aggregations to the server for later visualization. Input: name (string) for the data table and data (array of objects) representing the table. Returns: a success message.

visualize_data

Visualize a saved data table using Vega-Lite syntax. Input: data_name (string) for the data table and vegalite_specification (string) containing the JSON Vega-Lite spec. Returns: a text artifact with the complete spec and data if output_type is text, or a base64 PNG image if output_type is png.