Every chart type is its own tool with its own description of what it is for, so the decision between a column chart and a histogram is made where the data is being described rather than inside a charting library's API. Since the data goes in on the call, anything the model can put in a table it can also plot.
27 generators, one per output type: statistical charts (line, column, bar, area, scatter, histogram, boxplot, violin, radar, funnel, waterfall, sankey, treemap, venn, liquid, pie, word cloud, dual axes), diagrams (flow, mind map, fishbone, organization, network graph), three kinds of map, and a spreadsheet.
- A chart per shape of question — trends over time, comparisons between categories, distributions, proportions, correlations and flows
- Diagrams that are not data plots: flow charts, mind maps, fishbone cause analysis, organization charts and network graphs
- Maps: administrative districts, pinned points with popups, and a planned route
- A spreadsheet from tabular data, which renders as a pivot table when rows or values are supplied
- Title, axis titles, width, height, theme and style set per call, with stacked or grouped variants where the chart type supports them
Npx on your PATH. Each call carries its data inline, so nothing connects to a database — the numbers come from the conversation.
One command — npx -y @antv/mcp-server-chart