Two weeks ago at their Dev Days, OpenAI unveiled ChatGPT apps, interactive widgets that render inside ChatGPT using the Apps SDK and MCP. Building them using their template repository was surprisingly painful: every time you make a change on your React widget, you need to rerun the entire build pipeline to get fresh JS and CSS assets. The dev feedback loop was pretty terrible.
That’s why I built a TypeScript starter that gets you up and running in under a minute. Clone, npm install, npm run dev, paste your ngrok URL into ChatGPT, and you're iterating on widgets with Hot Module Reload in the ChatGPT interface, just like building a regular web app. When you're ready to ship, the production build pipeline deploys instantly to Alpic or any other PaaS.
What's included: - Vite dev server with Hot Module Reload (HMR) piggy-backed on your MCP server Express server - Skybridge framework: an abstraction layer I built on top of OpenAI's skybridge runtime that maps MCP tool invocations to React widgets, eliminating manual iframe communication and component wiring. - Production build pipeline: one-click deploy to Alpic.ai (with bundling, hosting, auth, MCP-specific analytics included) or elsewhere - No lock-in: uses the official MCP SDK, works with OpenAI's examples
Quick start:
git clone https://github.com/alpic-ai/apps-sdk-template
cd apps-sdk-template
pnpm install && pnpm run dev
ngrok http 3000
# Add https://your-url.ngrok-free.app/mcp to ChatGPT Settings → Connectors
Feedback welcome, especially if you hit rough edges or want specific features!
Starter Kit GitHub → https://github.com/alpic-ai/apps-sdk-template
Skybridge Github → https://github.com/alpic-ai/skybridge
