Microsoft Research releases Flint, a visualization language for AI-driven chart creation

Microsoft Research introduced Flint, an open-source visualization language intended to bridge the gap between compact chart specifications and polished visualizations. The project provides a compact, human-editable specification format that AI agents can reliably use to generate expressive, visually refined charts without requiring verbose, low-level parameters.

Flint leverages semantic data types—such as YearMonth, Profit, or Country—to guide the compiler in choosing appropriate scales, baselines, formatting, and color schemes. This approach reduces the need for users to manually configure fragile, library-specific details that are prone to error, especially in agentic workflows.

A single Flint specification can compile to multiple visualization backends—including Vega-Lite, Apache ECharts, and Chart.js—without rewriting the chart from scratch. This portability allows users to target the backend whose capabilities best fit the intended visualization while keeping the high-level intent unchanged.

The project includes the flint-chart library and the flint-chart-mcp server, enabling agents to create, validate, and render charts directly within chat or coding environments. This design is aimed at improving reliability and inspectability when LLMs and AI agents take on visualization tasks.

In a research comparison against a baseline that prompted models to generate full Vega-Lite specifications directly, Flint reduced the burden on models to produce low-level, brittle code. Instead, models focus on higher-level intent—such as identifying semantic types—which the compiler then translates into polished, backend-native visualizations.