Your LLM's new chart department: 26 tools, zero design skills
An MCP server that turns conversational data requests into AntV visualizations without you picking colors or chart types.

What it does This is a Model Context Protocol server that exposes 26+ chart-generation tools to any MCP client (Claude, Cursor, VS Code, etc.). You describe your data or ask for a visualization; the server returns a rendered chart image via AntV’s rendering pipeline. It also offers a separate “skill” mode that auto-selects chart types if your IDE supports it.
The interesting bit The project is essentially a well-organized remote control for AntV’s charting engine, wrapped in MCP’s tool-calling protocol. The geographic charts (district, path, pin maps) are powered by AMap and locked to China coverage — a specific regional dependency that’s unusual for a broadly distributed open-source tool.
Key highlights
- 26+ discrete tools: from boxplots and violin plots to fishbone diagrams, Sankey flows, and pivot tables
- Three transport modes: stdio (default for local MCP clients), SSE, and Streamable HTTP
- Private deployment path: swap in your own
VIS_REQUEST_SERVERpointing to a self-hosted GPT-Vis-SSR instance - Docker compose setup included for server deployments
- Tool filtering via
DISABLED_TOOLSenv var if you want to limit what your LLM can summon - Dify marketplace plugin available for no-code workflow builders
Caveats
- Geographic map tools require AMap and only support China; private deployments lose these entirely
- The default chart generation service is free but hosted; you’ll need Alipay for the optional generation-records feature
- README is thorough on setup but leaves the actual chart configuration schema mostly implicit — you’ll learn by doing or reading source
Verdict Worth wiring up if you’re already living in MCP-land and tired of LLMs hallucinating ASCII charts. Skip it if you need global maps or want fine-grained control over every visual parameter; this prefers convention over configuration.