I built DYPAI a backend platform where you describe what you need to your AI assistant (via MCP) and it builds the entire backend for you.
Instead of writing API routes or edge functions, your backend is a composition of nodes: database queries, Stripe payments, emails, Slack, WhatsApp, AI agents, logic – all connected as flows and exposed as production API endpoints.
One prompt, one endpoint: validate input → check permissions → query DB → transform data → send email → notify Slack → respond. No code. No deploy. Just works.
You can build from Cursor/Claude via MCP, see everything on a visual canvas, or use the REST API and TypeScript SDK directly.
It's in public beta so rough edges are expected – that's exactly why I'm sharing it here.
Demo video: https://www.dypai.ai/
Looking for feedback. Happy to answer anything.