Skupost — AI Social Content SaaS
Web Development
Solo SaaS built for online sellers: paste one product link and get a week of Reels, carousels, Stories, captions, and ads. Live on skupost.com.
- Next.js
- TypeScript
- Tailwind CSS v4
- Supabase
- Polar
- Vercel
- Upstash Redis
- Sentry
Context
Small online sellers know their product but lose hours every week deciding what to post. They needed a tool that reads a real product page and returns ready-to-copy posts that stay true to the listing — no invented prices or claims — with a free demo that converts to a paid plan without a sales call.
My role
Web Development
Architecture
I designed and built the whole product alone: Next.js App Router with strict TypeScript and Tailwind v4, Supabase (Auth with email/password and Google OAuth, Postgres with RLS on every table), deployed on Vercel. A product-extraction engine reads public listings (Shopify, Allbirds, IKEA and more) behind SSRF protection. A two-stage LLM pipeline writes 13 posts in three selectable voices, validated with Zod and guarded against hallucinated facts — prices come only from the listing. An anonymous landing demo uses hashed-IP daily limits and Upstash Redis rate limiting, then lets the visitor claim the result after sign-up. Subscriptions run on Polar with signature-verified, idempotent webhooks and monthly usage quotas. Shareable public result pages, history with search, settings, a gated admin dashboard, and an internal Channel Finder tool that scores communities and drafts launch copy. Error monitoring with Sentry; SEO and AEO with JSON-LD, sitemap and llms.txt.
Media
