Why FastSvelte
For the last 2 years, I’ve shipped multiple AI SaaS projects with SvelteKit + FastAPI and backend + SPA pattern has worked really well. I decided to make this a SaaS starter kit and become my first customer for my future projects and built FastSvelte.
I wasn't expecting any sales but sold 3 licenses and got positive feedback from customers. That motivated me to build FastReact (https://fastreact.dev) due to React's popularity.
I poured my 15+ years of development experience into this. Both FastSvelte and FastReact share the same FastAPI backend.
Recenty, I've added AI-SaaS related features like token metering and billing. Also a fully working, sample AI-pilot feature.
I am really proud of both projects. I've spend around 8 months so far on building these.
Some technical details: Backend does the heavy-lifting. Frontend is designed as a presentation layer with minimal business logic. Backend uses a 3-layered (data, service, api) architecture and dependency injection, similar to Spring Boot but. I really gave a lot of thought on how to organize the backend. As a result, Claude Code can understand the codebase really well and and easily add a new feature by applying the existing pattern. Docs contain much more details that are free to access: https://docs.fastsvelte.dev/
Really like the thinking behind this. A lot of AI SaaS projects start by moving fast but later struggle with architecture as features like billing, usage tracking, and AI workflows grow. Separating the backend into clear layers and keeping the frontend as a presentation layer seems like a strong foundation for long-term maintainability.
I'm curious — with AI coding tools becoming part of the development workflow, do you think architecture patterns and code organization are becoming even more important because they directly affect how well agents can understand and extend a codebase?