Autos Alemanes · Costa del Sol
An AI-assisted platform to find, score and track profitable German used cars for resale in Spain
The Problem
Finding the right used German car to resell on the Costa del Sol means hours and hours of manual searching: opening dozens of listings across several marketplaces, comparing prices by hand, spotting duplicates, and still not knowing if the deal leaves a real profit in Spain. I started this project to eliminate that time and to make the decision data-driven.
The Solution
A local web app that searches two German marketplaces — Kleinanzeigen and AutoScout24 — simultaneously, cleans the results, scores every car, and estimates the profit of reselling it in Spain.
How It Works
- 01
Search
Configurable filters — brand, model, city or postal code, year, mileage, price range, fuel type and gearbox. Both sources run in one sync.
- 02
Clean
Automatically removes electric cars, cargo vans, damaged vehicles and duplicate listings that appear on both marketplaces.
- 03
Score
Every car gets a 1–10 score. The user configures how much each factor weighs.
- 04
Analyze
On demand, compares the car with similar listings in Spain and calculates the estimated margin.
Smart Sync
The list refreshes automatically when the app opens if the data is stale (the user sets how many hours count as stale), or manually with one button. An optional daily scheduled sync is available for always-on servers. Before each run, the app waits for the network to be available and keeps the computer awake while it runs.
The Configurable Score
The score is deterministic and fully transparent — no AI involved, so it's free, fast and explainable. Weights are normalized and every user can tune them to their own strategy.
The price component: an expected price per model is learned from the whole database, adjusted for year and mileage (about 23% typical error). Cars priced well below that expectation score higher.
Profitability Analysis
One click on any car compares it with comparable listings on AutoScout24.es, adjusts prices for year and mileage, discards outliers, and calculates the margin after transport, taxes (including the Spanish REBU special regime for used goods) and other costs. The result is a pessimistic–optimistic range and a Buy / Doubtful / Discard verdict with a confidence level. All costs are editable. An optional AI-written report (Gemini or Claude, configurable) adds strengths, risks and what to inspect — it only receives the numbers already calculated, and runs once per request to keep costs under control.
- Line-by-line breakdown: expected sale price in Spain minus purchase price, transport, paperwork, registration tax, preparation, insurance and advertising — every line adds up exactly to the net margin.
- Comparable listings: real cars on the Spanish market, adjusted for year and mileage, with a confidence level (high / medium / low) based on how many comparables were found.
- Honest by design: with low confidence the tool never recommends 'Buy', and it flags when a listing's title contradicts its registered model.
Estimates are orientative, not tax advice.
You Control the Numbers
Every cost and threshold behind the profitability analysis is editable, so the tool adapts to each buyer's real situation instead of imposing generic assumptions.
Import costs
Transport to the Costa del Sol, registration paperwork (DGT fee, import ITV, plates), registration tax, preparation and small repairs, insurance and trade plates, advertising costs, and any other costs.
Buyer profile
The user chooses how they import — as a private individual or as a company — and the margin calculation adapts, including Spain's REBU special regime for used goods.
Negotiation and resale assumptions
Expected discount on the asking price when buying, and the discount applied to Spanish listing prices when selling, because listings are asking prices, not final prices.
Decision thresholds
The user defines the minimum net margin and return on investment to get a "Buy" verdict, and the margin below which a car is "Discard". Everything in between is "Doubtful".
Estimates are orientative, not tax advice. They don't include income tax, corporate tax, the buyer's transfer tax or unforeseen costs.
Deal Tracking (CRM)
Each car can be marked with a status, notes and an appointment date, so the whole sourcing pipeline lives in one dashboard.
Engineering Highlights
Fault-tolerant scrapers
Retries and per-source isolation — one failing site never stops the others; partial data is preserved.
Smart duplicate detection
Across marketplaces, with conservative rules so a different car is never hidden by mistake.
Respectful scraping
Pauses between requests and page limits.
Ethical by design
mobile.de blocks automated browsers; I documented it and chose not to circumvent the block.
Prompt-injection-safe AI
Seller text is isolated and treated as untrusted; the AI response is validated.
Background sync
Live status, per-source freshness indicators and delisted-ad detection.
Deleted-listing detection
A removed ad can still look alive on desktop. The scraper checks the mobile version, where deletion is shown, and treats ambiguous errors as "still active" so a block never hides a good car.
Title vs. model sanity check
Sellers sometimes register the wrong model (e.g. an A180 listed as A160). The app detects the contradiction, lowers the confidence and never issues a "Buy" verdict on unreliable data.
Tech Stack
Built with Claude Code as a programming assistant — I defined the requirements, reviewed the code and validated every result against the live sites.