Hi HN, I'm Matt, solo developer behind Stellify.

The core idea is that code stored as structured JSON rather than text files gives AI much better context to work with. Structured storage lets us attach metadata capturing intent and relationships, and create feedback loops where AI updates that metadata as it works.

It uses a fixed stack composed of Laravel on the backend, a custom framework I built that brings Laravel's fluent API style to the frontend (whilst offering adaptors for Vue/ React/ Svelte) and various libraries that cover common requirements such as authentication. If something's missing, you can request it and I'll look at adding it.

Once you're ready to ship, the platform assembles the data back to source files so that you can export or deploy your code. There's no lock-in.

The Stellify platform provides a fully featured interface builder and code editor. There's also an open source MCP server that enables integration with Claude and other AI tools.

Happy to answer any technical questions about the architecture - there's a free tier if you want to poke around and try it out for yourself.