Case study · 2026 · Custom software · Web · AI
EU auctions are full of cars you can flip in Hungary - if you can find them fast enough.Built an always-on scraper that does the lookup and the maths, and pushes the deals to Telegram.
AutoImport scrapes Openlane Buy-Now lots, matches each against actual hasznaltauto.hu listings (canonical make/model + body type + powertrain + km/year windows), and computes a margin that already includes transport, buyer fees, and Hungarian registration tax. New deals - and only deals that clear costs - reach a private Telegram channel via @autoimportszoftverbot as a swipeable photo carousel with a /leaderboard command and auto-login buttons straight into the platform.
- Next.js
- TypeScript
- Postgres
- Playwright
- Telegram
Inside the build
Overview
- EU→HU
- Arbitrage route
- Hourly
- Deal refresh
- Landed
- Cost basis
- Telegram
- Deal delivery
AutoImport scrapes EU B2B car-auction Buy-Now lots, matches each lot against real Hungarian used-car listings, and surfaces only vehicles where the verified resale margin clears every cost - transport, buyer fees, Hungarian registration tax. New deals push to a private Telegram channel as a swipeable photo carousel.
What shipped
What it does
- Live EU auction scraper (Openlane Buy-Now)
- Comp-matched against Hungarian retail listings
- Landed-cost margin · transport + fees + reg tax
- Telegram alerts with a swipeable carousel
- Dealer dashboard · deals, borderline, history
The problem
- Dealers do this lookup in Excel at midnight and miss the deals
- A 'cheap price abroad' isn't a deal until landed cost is in
- New lots come and go in hours · polling five sites by hand is hopeless
Why it matters
- Every alert is a deal that already clears every cost
- New lots reach you within an hour
- The borderline cases are visible - not just the obvious wins
How it shipped
- 01 · BRIEF
Pin the auction source and the Hungarian comp source that actually drive margin.
Workshop with the dealer to confirm Openlane Buy-Now as the lot source and hasznaltauto.hu as the comp source - and to lock the landed-cost formula (auction + buyer fee + transport + Hungarian reg tax by fuel / cc / emission / year).
- 02 · BUILD
Scraper, comp-matcher, landed-cost model - and a Telegram bot that swipes.
Per-source scrapers normalise lots into a single internal schema. The matcher restricts comps by canonical make/model + body type + powertrain + km/year windows, then aggregates with time-decay weighting + a conservative low-end anchor (cheapest realistic HU comp, not the median) so margins are realistic floors.
- 03 · SHIP
Dashboard live, Telegram carousel posting deals - every hour.
Dealer dashboard exposes actionable / borderline / all tabs with a one-click refresh. The Telegram bot drops new deals into a private channel as a swipeable photo carousel with a /leaderboard for top-by-score; auto-login buttons take channel members straight into the platform.
Stack
Scraper
Per-source feeds, normalised
Openlane Buy-Now lots + hasznaltauto.hu retail listings on a single internal schema · sold/expired lots swept inactive each pass.
Matcher
Comp-anchored margin, not vibes
Canonical make/model + body + powertrain + km/year windows · time-decay weighting + low-end anchor so the margin is a floor, not a fantasy median.
Dashboard
Actionable · borderline · all
Dealer-grade UI with the three tabs that match how a buyer actually reads their inbox · one-click refresh.
Telegram bot
@autoimportszoftverbot · carousel + commands
The bot posts every new deal to a private channel as one swipeable photo with previous and next arrows (file_id-cached for instant image swap) and a /leaderboard command for the top-by-score view. 'Megnyitás a felületen' and 'Összes hirdetés' buttons use a magic auto-login link so channel members open the platform without a password.
Case study
“We were checking five auction sites with a spreadsheet at midnight and still missing the good lots. The studio built a system that watches the auctions for us, compares each car to the actual Hungarian used-car market with the import tax and shipping already in, and sends the real deals to our Telegram. We open the app from the message, the deal's already calculated. We've stopped guessing.”
More work
Next project
AxisFit