Overview
A Model Context Protocol (MCP) server implementation that provides the LLM an interface for visualizing data using Vega-Lite syntax. The server exposes two core tools: save_data for persisting a data table under a given name, and visualize_data for rendering the saved data as a Vega-Lite visualization by supplying the data_name and a Vega-Lite specification in JSON string form. Output behavior is configurable via the --output_type flag: textual mode returns a success message plus an artifact containing the full Vega-Lite spec with embedded data, while PNG mode returns a base64-encoded image through the ImageContent container. The README also shows how to wire this server into Claude Desktop by adding an MCP server entry with an output_type option, enabling prompts to request and receive visualizations as either specs or rendered images.
Features
save_data
Save a table of data aggregations to the server for later visualization. Input: name (string): data table name; data (array): array of objects representing the data table. Returns: a success message.
visualize_data
Visualize a saved data table using Vega-Lite. Input: data_name (string) and vegalite_specification (string) containing a JSON Vega-Lite spec. Returns: text output with a success message and an artifact containing the complete Vega-Lite spec with data, or a PNG image via ImageContent when output_type is png.
Output type configurability
Configure output_type to text or png to receive either the Vega-Lite spec (with data) or a base64-encoded PNG image of the visualization.
Who Is This For?
- Data Scientists:Visualize saved data tables via Vega-Lite in LLM-assisted workflows.
- LLM Developers:Expose data visualization capabilities to language models through MCP endpoints.
- Claude Desktop users / BI engineers:Integrate visualization capabilities into workflows using Claude Desktop config with --output_type settings.
