Flint: A Visualization Language for the AI Era

Please visit: Flint Project Site | MCP Server Guide

Flint is a visualization intermediate language that lets AI agents create expressive, polished visualizations from simple, human-editable chart specs. Instead of asking agents or developers to tune verbose chart configuration details such as scales, axes, spacing, labels, and layout, the Flint compiler derives optimized chart settings from the data, semantic types, chart type, and encodings. The result is a compact chart specification that agents can produce reliably, people can edit directly, and multiple backends can render as native Vega-Lite, ECharts, or Chart.js specs.

This repo contains two main components:

• flint-chart: a JavaScript/TypeScript library that compiles the same Flint input into Vega-Lite, ECharts, or Chart.js specs.
• flint-chart-mcp: an MCP server that lets agents create, validate, and render charts directly from a chat or coding environment.

Features

• Semantic chart specs. Flint captures what each field means using 70+ semantic types such as Rank, Temperature, Price, or Country.
• Automatic layout. Flint adapts sizing, spacing, labels, marks, and legends to the data cardinality, chart design, and canvas constraints.
• Multiple backends. Compile one input to 30+ chart types across Vega-Lite, ECharts, and Chart.js, with more to come soon.
• Agent-ready chart authoring. The MCP server gives agents Flint tools and chart guidance so they can choose a template, validate it, and open an interactive chart view in MCP-capable clients.

_{Flint turns compact chart specs into backend-native specs and rendered visualizations.}

Install

# Use Flint in your JavaScript/TypeScript codebase
npm install flint-chart

# For agents and MCP clients
npx -y flint-chart-mcp

_{Python package: to be released. The current Python port is a source-only preview in this repo.}

Use Flint As A Library

Every backend accepts the same ChartAssemblyInput and returns the target library's native spec object.

import { assembleVegaLite } from 'flint-chart';

const spec = assembleVegaLite({
  data: { values: myData },
  semantic_types: { weight: 'Quantity', mpg: 'Quantity', origin: 'Country' },
  chart_spec: {
    chartType: 'Scatter Plot',
    encodings: { x: { field: 'weight' }, y: { field: 'mpg' }, color: { field: 'origin' } },
    baseSize: { width: 400, height: 300 },
  },
});
// → a ready-to-render Vega-Lite spec

Swap the backend without changing the input shape:

import { assembleECharts, assembleChartjs } from 'flint-chart';

const echartsOption = assembleECharts(input);
const chartjsConfig = assembleChartjs(input);

See the API reference, backend references, and live editor for more library examples.

Use Flint As An MCP Server

Install flint-chart-mcp as a Model Context Protocol server when you want an agent to create charts in the same conversation where the question starts. It can open an interactive chart view, return static PNG/SVG output, or produce backend-native chart specs.

For setup, start with the Flint MCP project page. It includes client configuration, usage examples, and links to deeper references.

MCP calls let agents embed rows directly as data.values, or read local JSON, CSV, or TSV files by data.url. For agent workflows without MCP, use the standalone agent skill.

Repository overview

flint-chart/
├── packages/
│   ├── flint-js/          npm package `flint-chart` (TypeScript)
│   │   └── src/
│   │       ├── core/      semantics, layout, decisions, shared types
│   │       ├── vegalite/  Vega-Lite backend
│   │       ├── echarts/   ECharts backend
│   │       ├── chartjs/   Chart.js backend
│   │       └── test-data/ fixtures + generators (drive tests and the gallery)
│   ├── flint-py/          Python port preview (package to be released)
│   └── flint-mcp/         npm package `flint-chart-mcp` (MCP render server)
├── site/                  Vite + React demo: landing, gallery, editor, docs
├── agent-skills/          fallback copy of the MCP-served agent skill
├── shared/test-data/      JSON fixtures shared across JS + Python
└── docs/                  architecture and design documents

Documentation

The project site is the main entry point for examples, the live editor, and concept docs. For source-level references, start with the API reference, the Flint MCP project page, or the Development guide.

Contributing

Contributions are welcome! See .github/CONTRIBUTING.md and the Development guide.

git clone https://github.com/microsoft/flint-chart
cd flint-chart
npm install            # root workspaces: packages/flint-js + flint-mcp + site

npm run typecheck      # typecheck packages/flint-js + packages/flint-mcp
npm run test           # Vitest (packages/flint-js + packages/flint-mcp)
npm run build          # build packages/flint-js + packages/flint-mcp
npm run site           # demo site (gallery + editor) at http://localhost:5274/

Node 18+ is required. The demo site aliases flint-chart to packages/flint-js/src, so library edits hot-reload in the gallery and editor without rebuilding dist/.

We especially welcome contributions that add new chart templates or new rendering backends.

This project has adopted the Microsoft Open Source Code of Conduct. See SECURITY.md to report vulnerabilities.

Contributor

Flint is built by Microsoft Research in collaboration with the IDEAS Lab, Renmin University of China. We welcome you to join us — see Contributing to get involved.

A research paper describing Flint is coming soon.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos is subject to those third parties' policies.

License

MIT © Microsoft Corporation