SellerOS — the platform that runs my store
Not a demo. A real system that closes real tasks every day, on algorithms I designed as the expert.
A #1 store is 5–6 processes running on non-trivial algorithms: what to ship where, what to reorder from China,
how to forecast stock three months out, how to run ads.
These only work if they're built on real expertise. I ran this store by hand for years — then
turned my read of the market into algorithms. The platform is that expertise, packed into a system that runs itself.
What to ship, and where
Every week: which products to which regional warehouses. That's 500+ product-warehouse pairs,
each scored on five criteria — impossible by hand or in spreadsheets. We barely did it, and left ~15% of profit
on the table. The platform scores every pair and picks the most profitable shipments.
How much to reorder from China, and when
It looks like you can eyeball it — but order accuracy drives your annual return on cash: every
extra unit in a warehouse is frozen capital. You have to nail it across 30–40 products at once — forecast sales,
track live stock, plan inventory months ahead with seasonality. The platform does it for me.
A store that runs itself
Prices, stock, ads, warehouse check-ins — the platform watches it all daily and pushes
ready-made decisions to Telegram. That's how a huge store runs on two people, with me out of the day-to-day.
Designed by me, built in Claude Code (the agent that writes the code).
Node.jsSQLiteOzon Seller APITelegram Bot APIcron
75 API endpoints · 19-table database · nightly pipelines with retries around a rate-limited API.
