Microsoft Research has introduced Flint-interactive, a language for creating interactive data visualizations on top of Flint. The research organization says the system is designed to help AI agents and users produce charts that people can explore directly through actions including clicking, filtering, zooming, highlighting and selecting data across multiple views.

Flint-interactive uses high-level semantic events and update operators rather than requiring users or agents to write low-level visualization code. According to Microsoft Research, its compiler translates interactions such as clicks and drags into events that identify chart elements and data, then applies requested updates to the visualization. Supported updates include changing styles, adding annotations and adjusting the viewport.

The system can also connect charts with text, tables, widgets and application state. Microsoft Research gives examples including brushing one view to highlight related groups in another, selecting a chart element to pass context into a follow-up chat turn, and using a table selection to update a chart. Developers can define custom handlers for interactions beyond Flint-interactive’s presets.

Microsoft Research says Flint-interactive is available for use in chat or canvas environments through Flint-MCP and MCP Apps, and can also be used to build dashboards, data stories and other applications through Flint Interactive Canvas. The organization presents the compiler-based approach as a way to reduce the development and debugging work involved in coordinating interaction and rendering logic.