Building an AI Agent That Finds the Best Deals on eBay and Jawa

Step by step

1. 1

   Define the search

   Write down the exact product, acceptable condition and maximum total price before building anything.

2. 2

   Fetch structured listings

   Fetch each marketplace's public search page and extract listings into a fixed schema rather than passing raw HTML to the model.

3. 3

   Score deals in code

   Calculate price plus shipping and compare it with the median of comparable listings. Arithmetic belongs in code, not in the model.

4. 4

   Let the model summarise

   Pass the top candidates to the model to explain trade-offs and send you a short report.

Example

from statistics import median

def best_deals(listings: list[dict], max_total: float) -> list[dict]:
    totals = [l["price"] + l.get("shipping", 0) for l in listings]
    typical = median(totals)
    deals = [
        {**l, "total": t, "vs_median": round(t / typical, 2)}
        for l, t in zip(listings, totals)
        if t <= max_total
    ]
    return sorted(deals, key=lambda d: d["total"])[:5]

Frequently Asked Questions