eBay sold comps API: build pricing on real sold data

Most of the teams pulling sold data from CompSniper are builders, not casual sellers: pricing apps, repricers, inventory spreadsheets, card and video game price trackers, buy-box tools for stores and pawn shops, and AI agents that answer "what is this worth?". They search across electronics, video games, sports cards, Pokemon, coins, toys, LEGO and sneakers, often thousands of times a day.

What they need from a sold data API is simple: real completed sales, fast, in a stable JSON shape, with the math already done. That is what this guide builds.

Get a free key at compsniper.com/signup, set COMPSNIPER_API_KEY, and turn sold comps into a price suggestion:

import os
import requests

API = "https://api.compsniper.com/v1/scrape"
HEADERS = {"Authorization": f"Bearer {os.environ['COMPSNIPER_API_KEY']}"}

def sold_comps(keyword, site="ebay.com", **filters):
    r = requests.get(API, headers=HEADERS, timeout=75, params={
        "keyword": keyword, "ebaySite": site, "sold": "true",
        "count": 240, "relevance": "true", **filters})
    r.raise_for_status()
    return r.json()

def suggest_price(keyword, **filters):
    s = sold_comps(keyword, **filters).get("summary") or {}
    if not s or s["count"] < 5:
        return {"keyword": keyword, "price": None, "reason": "not enough sold comps"}
    return {"keyword": keyword, "price": s["median"], "low": s["p25"],
            "high": s["p75"], "comps": s["count"], "currency": s["currency"],
            "confidence": "high" if s["count"] >= 20 else "low"}

print(suggest_price("nintendo switch oled"))
# {'keyword': 'nintendo switch oled', 'price': 180, 'low': 159.99,
#  'high': 219.99, 'comps': 181, 'currency': 'USD', 'confidence': 'high'}

The output numbers are from a real ebay.com search collected on October 2, 2026. Prefer JavaScript? The JavaScript and TypeScript guide has the same request.

The rows behind the summary are in items: sold price, shipping, sold date, condition, seller and a bestOfferAccepted flag. Show users a few of them next to your suggested price; seeing real sales builds more trust than a single number.

Loop over your items and write the results to a spreadsheet. One request per item, whatever the number of comps:

import csv
import time

inventory = ["nintendo switch oled", "pokemon japanese pikachu 291 psa 10",
             "lego 75192 millennium falcon", "sony wh-1000xm5"]

with open("prices.csv", "w", newline="") as f:
    writer = csv.DictWriter(f, fieldnames=["keyword", "price", "low", "high",
                                           "comps", "currency", "confidence", "reason"])
    writer.writeheader()
    for keyword in inventory:
        try:
            writer.writerow(suggest_price(keyword))
        except requests.HTTPError as err:
            # 429/503: wait and retry later. Failed requests are not charged.
            writer.writerow({"keyword": keyword, "reason": str(err)})
        time.sleep(1)

For bigger runs, bulk search takes up to 20 keywords at once and the Card Batch API is built for trading-card lists. Building inside Claude or Cursor? The MCP server gives an AI agent the same sold comps.

• Search the exact item. Model numbers, set and card number, storage and size belong in the keyword.
• Filter condition. itemCondition=used or new, or a specific conditionId.
• Leave relevance=true on. It removes accessories, parts, lots and wrong models before the summary. Our 100-product cleaning study shows where it is strong and where to double-check.
• Pick the right marketplace. Set ebaySite for your user's country so prices come back in their currency.
• Respect small samples. Our sample-size guide explains when a median is trustworthy.

Written by CompSniper's founder. Checked October 3, 2026.