I built an AI tool that turns photos into eBay/Etsy listings in seconds
Hey IH! 👋
I've been working on Underpriced AI - an app that helps resellers (thrift flippers, eBay sellers, etc.) photograph items and get instant AI-powered valuations + ready-to-post listings.
The Problem
Resellers spend hours researching what their items are worth and writing listing descriptions. It's tedious and most people underprice their stuff because they don't know what they have.
The Solution
Snap a photo → Claude AI identifies the item, estimates value, and generates an optimized listing title + description. Copy/paste to eBay or Etsy.
Tech Stack
- Next.js 16 + React 19
- PostgreSQL + Prisma
- Claude API (Anthropic) for image analysis
- AWS S3 for image storage
- AWS SES for transactional emails
- Stripe for subscriptions
- Capacitor for iOS/Android apps
Where I'm At
- MVP complete with web + mobile apps
- 5 early users testing
- iOS app in TestFlight, Android ready for Google Play
- Just shipped welcome emails and AWS Secrets Manager integration
What's Next
- Direct eBay/Etsy API posting (no more copy/paste)
- Cross-posting to multiple marketplaces
- Profit tracking
The Competition
There are a bunch of apps in this space (Snap2List, SnapList, ListerMate, eProfit) but most are:
- eBay-only
- Mobile-only
- Using basic AI models
I'm betting that Claude's vision capabilities give better item identification and more accurate pricing than competitors.
Would love feedback from anyone who resells or knows the space!
Validating an idea to help professionals reply safely to difficult work messages
The "underpricing problem" is a great pain point to target - most resellers genuinely don't know what they're leaving on the table. Using Claude for the image analysis is smart; vision models have gotten significantly better at identifying obscure items and condition nuances.
A few things I'd be curious about:
Accuracy validation - How are you measuring pricing accuracy? With vintage/collectible items, the spread between "sell it tomorrow" prices and "wait for the right buyer" prices can be 3-5x. Are you surfacing that range or picking a point estimate?
Category-specific models - Do you see different accuracy patterns across categories? I'd guess electronics and standardized goods perform well, but handmade/vintage items might need more contextual training.
Feedback loops - Are you capturing actual sale prices to improve the model? That's the data moat that could separate you from competitors using the same underlying AI.
The direct marketplace API posting roadmap makes sense - that's where the real time savings compound. Good luck with the launch!
Hi, great product, I love the idea. Good luck with monetising.