🎨 Best DeepSeek Harness Design Plugin. The open-source Claude Design alternative. 🖥️ Local-first desktop app. 🖼️ Your coding agent becomes the design engine: prototypes, landing pages, dashboards, slides, images & video — real files, HTML/PDF/PPTX/MP4 export. 🤖 Claude Code / Codex / Cursor / DeepSeek Harness / OpenCode & 20+ CLIs via BYOK.
Generative UI demo for CopilotKit patterns
This playground shows three ways to build AI-powered interfaces with CopilotKit: static React UI from tool calls, MCP Apps in sandboxed iframes, and A2UI from declarative JSON. The frontend switches between agent setups and renders the matching UI pattern.
Builders who want to compare CopilotKit UI patterns in a running demo.
You can see how the same chat flow produces different kinds of generated UI.
What it does
Static GenUI
Renders pre-built React components like weather, stock, and approval cards from frontend hooks.
MCP Apps
Serves HTML and JavaScript apps from an MCP server and shows them in sandboxed iframes.
A2UI
Renders declarative JSON UI from a Python A2A agent through the `A2UIRenderer`.
Agent switching
Uses `CopilotKitProvider` and a sidebar chat to switch between the demo modes.
How to get it
- 1Create a .env file
OPENAI_API_KEY=sk-your-key-here MCP_SERVER_URL=http://localhost:3001/mcp A2A_AGENT_URL=http://localhost:10002
- 2Start all three services
# Terminal 1: MCP Server (port 3001) cd mcp-server && npm run dev # Terminal 2: Python A2A Agent (port 10002) cd a2a-agent && python -m agent # Terminal 3: Next.js Frontend (port 3000) npm run dev
README
Generative UI Demo
https://github.com/user-attachments/assets/79ead351-f63c-4119-9d28-9d604e7f8876
A generative UI playground showcasing the three types for building AI-powered user interfaces with CopilotKit.
Demo Overview
This demo demonstrates how different types of generative UI can be used to create rich, interactive AI experiences:
| Spec | Description | Use Case |
|---|---|---|
| Static GenUI | Pre-built React components rendered by frontend hooks | Weather cards, stock displays, task approvals |
| MCP Apps | HTML/JS apps served by MCP servers in sandboxed iframes | Flight booking, hotel search, trading simulator |
| A2UI | Agent-composed declarative JSON UI rendered dynamically | Restaurant finder, booking forms |
CopilotKit Features Used
- CopilotKitProvider - Main provider with agent switching
- CopilotSidebar - Chat interface component
- useRenderToolCall - Display-only tool rendering (WeatherCard, StockCard)
- useHumanInTheLoop - Interactive approval flows (TaskApprovalCard)
- A2UIRenderer - Renders A2UI declarative JSON from agent responses
- MCPAppsMiddleware - Bridges MCP server tools with UI resources
- BasicAgent - TypeScript agent for Static GenUI + MCP Apps
- HttpAgent - Connects to Python A2A backend for A2UI
Setup
Prerequisites
- Node.js 18+
- Python 3.11+
- OpenAI API key
Installation
# Clone and install dependencies
cd ui-protocols-demo
npm install
# Install MCP server dependencies
cd mcp-server
npm install
cd ..
# Install Python A2A agent
cd a2a-agent
pip install -e .
cd ..
Environment Variables
Create a .env file:
OPENAI_API_KEY=sk-your-key-here
MCP_SERVER_URL=http://localhost:3001/mcp
A2A_AGENT_URL=http://localhost:10002
Running the Demo
Start all three services:
# Terminal 1: MCP Server (port 3001)
cd mcp-server && npm run dev
# Terminal 2: Python A2A Agent (port 10002)
cd a2a-agent && python -m agent
# Terminal 3: Next.js Frontend (port 3000)
npm run dev
Open http://localhost:3000 to see the demo.
Usage
Static + MCP Apps Mode
Click the "Static + MCP Apps" tab to use:
- "What's the weather in Tokyo?" → Weather card
- "Get stock price for AAPL" → Stock card with sparkline
- "Open the calculator" → Interactive calculator app
- "Search for flights to Paris" → Flight booking workflow
A2UI Mode
Click the "A2UI" tab to use:
- "Find Italian restaurants nearby" → Restaurant list with booking
- "Show me Chinese food options" → Filtered results
- "Book a table for 4" → Interactive booking form
Architecture
Frontend (Next.js) ─────────────────────────────────────────────────────
├── Protocol tabs switch between agents
├── Static GenUI: useRenderToolCall, useHumanInTheLoop
├── MCP Apps: Automatic iframe rendering via middleware events
└── A2UI: A2UIRenderer for declarative JSON
│
┌─────────┴─────────┐
▼ ▼
"default" Agent "a2ui" Agent
BasicAgent + MCP HttpAgent → Python
Port 3001 Port 10002
Project Structure
ui-protocols-demo/
├── src/app/ # Next.js frontend
│ ├── page.tsx # Main page with agent switching
│ ├── theme.ts # A2UI theme configuration
│ ├── api/copilotkit/ # CopilotKit API route
│ └── components/ # React components
├── mcp-server/ # MCP Apps server
│ ├── server.ts # Tool registrations
│ └── apps/ # HTML app files
└── a2a-agent/ # Python A2A agent
└── agent/ # Agent modules
Learn More
Note: This project has been consolidated into the CopilotKit monorepo. The latest version lives at
examples/showcases/generative-ui-playground. Please open issues and pull requests in the main CopilotKit repository.
Files in the repo
- a2a-agent
- mcp-server
- public
- src
- .gitignore
- CLAUDE.md
- Dockerfile
- eslint.config.mjs
- next.config.ts
- package-lock.json
- package.json
- postcss.config.mjs
- railway.toml
- README.md
- tsconfig.json
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More other
"Vibe-Trading: Your Personal Trading Agent"
OpenHuman is an open source personal AI for Mac, Windows and Linux — local-first memory, agent orchestration, and deep research.

Your Personal AI Assistant; easy to install, deploy on your own machine or on the cloud; supports multiple chat apps with easily extensible capabilities.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
macOS video editor built for AI