🧩 AI Native Kit UI

Instantly turn MCP JSON results into polished ChatGPT App UIs. OpenAI Figma‑aligned React components purpose‑built for the ChatGPT Apps SDK.

🎪 Live Storybook · ⚡ Quick Start · 🧪 Examples · 🧱 Components

🔍 Overview

AI Native Kit UI bridges the gap between structured MCP JSON and beautiful, accessible UI for ChatGPT apps. Designed for the Apps SDK, it maps model/tool results directly to interactive, Figma‑aligned components, so you stop hand‑wiring UI and start shipping.

✨ What you get: Production‑ready React components, example patterns, hooks for Apps SDK, and a rich design‑token system.
🧭 Who it's for: Developers building ChatGPT Apps who want consistent, on‑brand UI without reinventing the wheel.

Why now? ChatGPT Apps (via the Apps SDK) expose results + UI metadata. This kit renders those results as native widgets with minimal code.

💡 Why You'll Love It

Developer Pain Point	How AINativeKit UI Helps
Manually converting JSON to UI	Native JSON -> UI mapping components streamline integration
Lack of ChatGPT‑specific components	Built specifically for the ChatGPT Apps SDK with optimized patterns
Inconsistent design & icons	Figma‑aligned tokens + typed icon library ensure visual consistency
Accessibility concerns	WCAG 2.1 AA mindful components with ARIA support
Poor developer experience	100% TypeScript, IntelliSense, Storybook docs

🚀 Core Features

🎯 Apps SDK Optimized: Components designed to work seamlessly with ChatGPT Apps SDK
🔄 JSON -> UI Mapping: Render structured MCP results with minimal glue code
🤖 AI-Tool Friendly: JSON schemas, component registry, and utilities for AI code generation
🎨 417 Figma‑Aligned Icons: Fully typed and tree‑shakeable
♿ Accessibility First: ARIA attributes & sensible focus management
🌗 Dark/Light Themes: Built‑in theme switching
🧩 Production‑Ready Blocks: Cards, lists, carousel, album, map, and more
🪝 OpenAI Hooks: useOpenAiGlobal, useWidgetState, useMaxHeight
📦 Tree‑Shakeable & Type‑Safe: Import only what you need

⚡ Quick Start

1) Install

npm install @ainativekit/ui
# or
pnpm add @ainativekit/ui
# or
yarn add @ainativekit/ui

2) Turn MCP JSON into UI

import { SummaryCard } from '@ainativekit/ui';
import '@ainativekit/ui/styles';

// Example MCP/tool JSON from your backend
const restaurantData = {
  title: "Little Nona's",
  subtitle: "1427 Via Campania",
  rating: "9.2",
  description: "A tiny, brick-walled trattoria tucked down a side street. The windows glow warm gold at night.",
  images: [ // Display up to 4 images
    { src: "https://images.unsplash.com/photo-1513104890138-7c749659a591?w=400", alt: "Pizza" },
    { src: "https://images.unsplash.com/photo-1621996346565-e3dbc646d9a9?w=400", alt: "Pasta" },
    { src: "https://images.unsplash.com/photo-1512621776951-a57141f2eefd?w=400", alt: "Salad" }
  ]
};

export function RestaurantListing() {
  return (
    <SummaryCard
      images={restaurantData.images}
      title={restaurantData.title}
      subtitle={restaurantData.subtitle}
      badge={restaurantData.rating}
      badgeVariant="success"
      description={restaurantData.description}
      buttonText="Add to Order"
      onButtonClick={() => navigate(`/restaurant/${restaurantData.id}`)}
    />
  );
}

3) Compose Flexible and Customizable Layouts

import { Card, Features } from '@ainativekit/ui';
import '@ainativekit/ui/styles';

export function DocumentCard() {
  return (
     <Card elevationLevel={1} interactive>
      <Card.Header>
        <Card.ChipGroup>
          <Card.Chip variant="neutral" size="sm">AINativeKit UI</Card.Chip>
          <Card.Chip variant="neutral" size="sm">Guide</Card.Chip>
        </Card.ChipGroup>
      </Card.Header>
      <Card.Image
        src="https://images.unsplash.com/photo-1499750310107-5fef28a66643?w=400&h=200&fit=crop"
        alt="AINativeKit UI library documentation"
      />
      <Card.Body>
        <Card.Title as="h3">Building AI-Native UIs</Card.Title>
        <Card.Description>
          Build modern, accessible UI with AINativeKit. Master reusable component patterns that scale.
        </Card.Description>
        <Card.Meta>
          <Features items={[
            { icon: 'clock', label: '10 min read' },
            { icon: 'calendar-today', label: 'October 30, 2025' }
          ]} iconSize={12} />
        </Card.Meta>
      </Card.Body>
      <Card.Footer>
        <Card.Actions align="start">
          <Card.ActionButton variant="primary">Explore Docs</Card.ActionButton>
        </Card.Actions>
      </Card.Footer>
    </Card>
  );
}

📚 Explore many more examples in Storybook -> https://www.ainativekit.com

🧪 Examples

Cards: Image, summary, list, discovery
Lists: Structured lists with rich content
Carousel: Horizontal scroll galleries
Album: Media gallery with fullscreen
Map: Location UI pattern with fullscreen

Tip: Copy any example from Storybook into your app and tweak the props.

🧱 Components

Core: Button (primary/secondary/tertiary/ghost) · Icon · Badge · Chip · Alert · Skeleton · Card

Patterns: Card variants · Carousel · List · Album · Map

🎨 Design System

Use consistent colors, typography, spacing, and elevation derived from OpenAI's Figma system.

import { colors, typography, spacing, elevation } from '@ainativekit/ui';

const style = {
  backgroundColor: colors.light.background.primary,
  padding: spacing[16],
  fontSize: typography.body.fontSize,
  boxShadow: elevation[1].shadow,
};

Icons:

import { Icon } from '@ainativekit/ui';
import { SettingsCog, Terminal, Star } from '@ainativekit/ui/icons';

// Preferred: Named icon components
<SettingsCog size="md" />

// Alternative: Dynamic icon by name
<Icon name="settings-cog" size="lg" />

🪝 OpenAI Hooks & Theme Management

Utilities to integrate with the ChatGPT Apps SDK runtime.

import {
  useOpenAiGlobal,
  useWidgetState,
  useMaxHeight,
  useTheme,
  useDisplayMode
} from '@ainativekit/ui';

function MyChatGPTWidget() {
  // Get theme (read-only in ChatGPT, controllable with ThemeProvider)
  const { theme, isControlledByChatGPT } = useTheme();

  // Access other OpenAI global values (reactive)
  const displayMode = useDisplayMode(); // 'inline' | 'pip' | 'fullscreen' | null
  const maxHeight = useMaxHeight(); // number | null

  // Or access any global property directly
  const locale = useOpenAiGlobal('locale'); // string | null

  // Manage persistent widget state
  const [state, setState] = useWidgetState({ count: 0 });

  return (
    <div
      className={theme === 'dark' ? 'dark-mode' : 'light-mode'}
      style={{ maxHeight: maxHeight ?? 600 }}
    >
      {/* your widget */}
    </div>
  );
}

Available Hooks:

useTheme() - Get current theme and optionally control it (see ThemeProvider below)
useDisplayMode() - Get current display mode (inline/pip/fullscreen)
useMaxHeight() - Get maximum height constraint for layout
useWidgetState(defaultState) - Persistent state across ChatGPT sessions
useOpenAiGlobal(key) - Access any window.openai property reactively
useWidgetProps(defaultProps) - Get tool output data

Theme Management

For development and standalone apps, use ThemeProvider to enable programmatic theme control:

import { ThemeProvider, useTheme } from '@ainativekit/ui';

function App() {
  return (
    <ThemeProvider defaultTheme="light">
      <MyApp />
    </ThemeProvider>
  );
}

function MyApp() {
  const { theme, setTheme, isControlledByChatGPT } = useTheme();

  return (
    <div>
      <p>Current theme: {theme}</p>
      <button
        onClick={() => setTheme?.(theme === 'light' ? 'dark' : 'light')}
        disabled={isControlledByChatGPT}
      >
        Toggle theme
      </button>
      {isControlledByChatGPT && (
        <p>Theme is controlled by ChatGPT</p>
      )}
    </div>
  );
}

Theme Behavior:

Inside ChatGPT: Theme is read-only (window.openai.theme), setTheme has no effect
Inside ThemeProvider: Theme is controllable, persists to localStorage
Standalone: Theme defaults to system preference or specified default

ThemeProvider Props:

defaultTheme - Initial theme ('light' or 'dark'), default: 'light'
storageKey - LocalStorage key for persistence, default: 'ainativekit-theme'
enableSystemTheme - Detect system preference, default: true

📘 TypeScript Support

AINativeKit UI exports comprehensive TypeScript definitions for the ChatGPT Apps SDK, giving you full autocomplete and type safety for window.openai.

Zero Boilerplate - Types Just Work

// Simply import the package
import '@ainativekit/ui';

// window.openai is now fully typed! ✨
const theme = window.openai?.theme; // 'light' | 'dark' | undefined
const toolOutput = window.openai?.toolOutput; // string | undefined
const displayMode = window.openai?.displayMode; // 'inline' | 'pip' | 'fullscreen' | undefined

// Full autocomplete for all methods
window.openai?.sendFollowUpMessage({ prompt: 'Show more' });
const result = await window.openai?.callTool('get_weather', { location: 'SF' });

Available Types

All ChatGPT Apps SDK types are exported:

import type {
  OpenAiGlobals,
  OpenAiApi,
  Theme,
  DisplayMode,
  UserAgent,
  SafeArea
} from '@ainativekit/ui';

// Use in your own type definitions
type MyWidgetProps = {
  theme: Theme;
  displayMode: DisplayMode;
};

Exported Types:

OpenAiGlobals - Complete window.openai globals interface
OpenAiApi - API methods (callTool, sendFollowUpMessage, etc.)
Theme - 'light' | 'dark'
DisplayMode - 'inline' | 'pip' | 'fullscreen'
UserAgent - Device type and capabilities
SafeArea - Safe area insets for mobile layouts
CallTool - Type-safe tool calling signature
And more... (see types.ts)

🤖 AI Tool Integration

AINativeKit UI is optimized for AI coding assistants through Context7 and runtime utilities.

Using with Context7

If you use Claude Desktop, Cursor, or other AI editors with Context7 support:

use context7 @ainativekit/ui

This will inject the latest component documentation directly into your AI's context, enabling:

Smart component suggestions
Accurate prop recommendations
Best practice guidance
Code generation with examples

JSON-to-Component Rendering

Dynamically render components from JSON/API data in your production app:

import { renderComponent, type ComponentConfig } from '@ainativekit/ui';

const config: ComponentConfig = {
  type: 'SummaryCard',
  props: {
    title: "Little Nona's",
    badge: "9.2",
    images: ["https://example.com/restaurant.jpg"],
    buttonText: "Add to Order"
  }
};

const card = renderComponent(config); // Renders <SummaryCard {...props} />

Runtime Utilities:

🔧 renderComponent() - Render from JSON config
✅ validateComponentConfig() - Validate component configs
📝 ComponentPropsMap - Type-safe prop definitions

AI Integration:

🤖 Context7: Documentation via "use context7" command
📚 Schemas: Can be generated with pnpm --filter @ainativekit/ui generate:schemas for external tooling (not used in runtime)
📖 Guide: See /docs folder for comprehensive documentation

🧰 Development

pnpm install          # install deps
pnpm storybook        # run interactive docs
pnpm test             # run tests
pnpm build            # build the library
pnpm lint             # lint

Package structure

@ainativekit/ui
├── /               # Components, tokens, hooks, utilities
├── /icons          # 417 icons as named React components
├── /tokens         # Design tokens only
└── /styles         # CSS styles

Repository (for documentation & tooling):
├── /docs           # Markdown documentation (Context7)
├── /schemas        # JSON schemas (generated on-demand, not committed)
└── /metadata       # Component registry (dev tooling)

✅ Compatibility

React ≥ 18
TypeScript ≥ 5 (recommended)
ChatGPT Apps SDK (preview)
Works with modern bundlers (Vite, Next.js, etc.)

🗺️ Roadmap (high‑level)

More first‑class MCP JSON -> UI mappers (tables, charts, forms)
Expanded widget patterns used commonly in ChatGPT apps
Theming API refinement + tokens export
Additional a11y audits

Have ideas? Please open an issue or PR!

🤝 Contributing

Contributions welcome! Please:

Star the repo to support visibility 🙌
Open an issue to discuss new features/bugs
Fork -> create a feature branch -> open a PR

See CONTRIBUTING.md for guidelines.

🔗 Links

Storybook: https://www.ainativekit.com
NPM: https://www.npmjs.com/package/@ainativekit/ui
GitHub: https://github.com/AINativeKit/ainativekit-ui
Issues: https://github.com/AINativeKit/ainativekit-ui/issues

🙏 Acknowledgments

Built for the OpenAI Apps SDK community. Inspired by ChatGPT App examples, OpenAI Figma design, Apple HIG, Material UI, Chakra UI, and Ant Design.

Made with ❤️ by and for ChatGPT App developers.

Stop wiring UIs manually, start shipping faster with AI Native Kit UI.

Name		Name	Last commit message	Last commit date
Latest commit History 44 Commits
.github/workflows		.github/workflows
assets		assets
eslint-local-rules		eslint-local-rules
packages/ui		packages/ui
.eslintignore		.eslintignore
.eslintrc.cjs		.eslintrc.cjs
.gitignore		.gitignore
.nvmrc		.nvmrc
.prettierignore		.prettierignore
.prettierrc		.prettierrc
CHANGELOG.md		CHANGELOG.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
context7.json		context7.json
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml
pnpm-workspace.yaml		pnpm-workspace.yaml
tsconfig.json		tsconfig.json

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Repository files navigation

🧩 AI Native Kit UI

🔍 Overview

💡 Why You'll Love It

🚀 Core Features

⚡ Quick Start

1) Install

2) Turn MCP JSON into UI

3) Compose Flexible and Customizable Layouts

🧪 Examples

🧱 Components

🎨 Design System

🪝 OpenAI Hooks & Theme Management

Theme Management

📘 TypeScript Support

Zero Boilerplate - Types Just Work

Available Types

🤖 AI Tool Integration

Using with Context7

JSON-to-Component Rendering

🧰 Development

✅ Compatibility

🗺️ Roadmap (high‑level)

🤝 Contributing

🔗 Links

🙏 Acknowledgments

About

Uh oh!

Releases 1

Packages

Contributors 2

Languages

License

AINativeKit/ainativekit-ui

Folders and files

Latest commit

History

Repository files navigation

🧩 AI Native Kit UI

🔍 Overview

💡 Why You'll Love It

🚀 Core Features

⚡ Quick Start

1) Install

2) Turn MCP JSON into UI

3) Compose Flexible and Customizable Layouts

🧪 Examples

🧱 Components

🎨 Design System

🪝 OpenAI Hooks & Theme Management

Theme Management

📘 TypeScript Support

Zero Boilerplate - Types Just Work

Available Types

🤖 AI Tool Integration

Using with Context7

JSON-to-Component Rendering

🧰 Development

✅ Compatibility

🗺️ Roadmap (high‑level)

🤝 Contributing

🔗 Links

🙏 Acknowledgments

About

Resources

License

Code of conduct

Contributing

Uh oh!

Stars

Watchers

Forks

Releases 1

Packages 0

Contributors 2

Languages

Packages