MCP Server Chart

What antv/mcp-server-chart MCP server does

antv/mcp-server-chart MCP server lets MCP-compatible AI applications request charts and other visual representations from supplied data. It is implemented in TypeScript and uses AntV for rendering. The repository describes chart generation and data analysis as its primary use cases, so it fits workflows where an agent must select or produce a visualization from a dataset or a visualization request.

The package also includes a spreadsheet generator. Depending on the supplied fields, that tool can render a regular table or a pivot-style cross-tabulation. Geographic tools are included for administrative regions, points of interest, and routes, but their geographic coverage is currently limited to China.

How it works

The server exposes separate MCP tools for individual visualization types. A client can call tools such as generate_bar_chart, generate_line_chart, or generate_scatter_chart, depending on the required representation. The available tools cover comparison, trends, distributions, proportions, relationships, flows, hierarchies, and process diagrams.

It can run with MCP stdio transport for local desktop clients. It also supports SSE and Streamable transports for an HTTP-accessible deployment. The documented local SSE endpoint is /sse on port 1122, while the Streamable endpoint is /mcp on the same port when launched directly with the corresponding transport option.

Setup and configuration

For a local stdio configuration, run the npm package with npx -y @antv/mcp-server-chart. The README shows this command in MCP configuration examples for desktop applications, including Claude, Cursor, and Cline. Windows configurations use cmd /c before the same npx invocation.

For an HTTP-style deployment, install the package globally with npm and start the CLI using --transport sse or --transport streamable. SSE is documented at http://localhost:1122/sse; Streamable transport is documented at http://localhost:1122/mcp. A Docker Compose deployment is also described in the repository's docker directory, with its own documented local endpoints.

Tools and capabilities

The listed capabilities include:

• Area, bar, column, dual-axis, line, pie, radar, and funnel charts
• Boxplot, histogram, violin, scatter, liquid, and word-cloud visualizations
• Treemaps, Sankey diagrams, Venn diagrams, network graphs, and mind maps
• Organization, flow, and fishbone diagrams
• District maps, path maps, and pin maps
• Regular spreadsheets and pivot-style tables

The geographic chart tools use the AMap service and are intended for data within China. An optional chart-visualization skill is documented separately for AI IDEs that support skills; it can help select among the chart types, but it is not the MCP server itself.

Limitations and notes

The repository does not describe this server as a database connector or data-ingestion system. Data must therefore come from the MCP client or another part of the agent workflow. Geographic visualization is restricted to China according to the README. The documented client examples cover Claude, VSCode, Cline, Cherry Studio, and Cursor, while other clients may require their own MCP configuration format.