DEV Community

James Walstonn
James Walstonn

Posted on

I built a free macOS app that automates eBay/Etsy/Amazon listing with on-device AI — no API keys needed

#ai #automation #productivity #showdev

Background

I run a small cross-border e-commerce operation — sourcing products from 1688.com and listing them on eBay, Etsy, AliExpress, and Amazon. For months I did this manually: open 1688, find a product, copy-paste the description, translate it, check the images, create the listing. Repeat 50 times.

So I built AutoStore to do it for me.

What it does

AutoStore is a free macOS app (+ Chrome extension) that automates the entire product sourcing pipeline:

  1. Discover products on 1688.com using 450+ blue-ocean search terms
  2. Filter brands — 548+ banned brand names checked against titles, descriptions, specs, and images via OCR
  3. Translate — runs Apple Foundation Models on-device, no API key, no cost
  4. Bulk-upload to eBay, Etsy, AliExpress, or Amazon via your already-logged-in Chrome session

The interesting technical bits

On-device AI translation

Instead of calling OpenAI or Baidu Translate, AutoStore uses Apple's on-device Foundation Models (macOS 15+). Zero cost per translation, works offline, and no data leaves the device. Product titles, descriptions, and specifications all get translated in batch.

Brand detection with OCR

Images from Chinese suppliers sometimes have brand logos or text watermarks baked in. AutoStore uses Apple Vision to OCR every product image and cross-reference against the brand blacklist. If "Nike", "Apple", or any of 548 other brands appears anywhere — in the spec sheet, title, image, or seller name — the product is skipped automatically.

Real Chrome automation

Most automation tools spin up a headless browser that platforms can detect and block. AutoStore instead drives your existing logged-in Chrome session via a Chrome extension. From the platform's perspective, it looks exactly like a human clicking through their own browser — because it is.

Local-first data

All data (product catalog, translation cache, brand list) lives in a local MySQL database. Credentials are stored in macOS Keychain. Nothing is sent to a server. PIPL compliant.

The pipeline in practice 

1688 discovery
    → brand title filter (isBannedBrand)
    → detail scrape + image download
    → image OCR check (Vision framework)
    → on-device AI translation
    → category enrichment
    → bulk CSV/Excel generation
    → Chrome extension upload to eBay/Etsy/AliExpress/Amazon
Enter fullscreen mode Exit fullscreen mode

It's free forever

AutoStore is free with no subscription and no commission. The target users are Chinese cross-border sellers who are underserved by existing tools (Helium 10, Jungle Scout, SellerSprite) — which are English-only, don't integrate 1688, and cost ¥300-1000/month.

Two versions: AutoStore Full (complete pipeline) and AutoStore Lite (manual import for sellers who already have their own sources).

Try it

🌐 spriterock.com

Happy to answer questions about the architecture — the Chrome extension ↔ macOS app IPC, the Ember-like OCR pipeline, or the MySQL schema.

Top comments (0)