Flint: A new visualization language letting AI agents create polished charts from simple specs

Flint addresses a key tension in data visualization: short specs using default settings often produce bland charts, while polished visuals require verbose, fragile parameters. This is especially problematic for LLM-driven agents that struggle with low-level spec details. Flint introduces a compact intermediate language that sits between simplicity and quality. The compiler infers optimal settings (scales, axes, spacing, color schemes, layouts) from semantic data types, chart type, and field encodings. It supports three major backends—Vega-Lite, ECharts, Chart.js—allowing one spec to render across platforms.

The open-source project ships two components: the flint-chart library for programmatic use, and the flint-chart-mcp server enabling AI agents to create, validate, and render charts directly in chat or coding environments. Semantic types (e.g., "period as YearMonth", "newUsers as Profit") guide design decisions like temporal parsing, axis formatting, color scales, and cell sizing. Layouts automatically adapt as cardinality or density changes, and the resulting backend-native specs are fully editable by humans. Flint promises to make AI-generated charts both reliable and beautiful.