eBay Sold Listings Scraper

eBay Sold Listings Scraper — sold comps API, no eBay key

Search eBay's sold and completed listings and get them back as rows: item id, title, sale price, currency, shipping cost, sold date, condition, buying format, seller location and image.

This is what you use to price something. What did this actually sell for, how many times, at what spread, in the last N days.

No eBay developer account, no OAuth, no browser.

Output

A real row, as it lands in the dataset:

{
  "ok": true,
  "_sample": false,
  "itemId": "396512340987",
  "title": "Nintendo Switch OLED Console White Joy-Con — Very Good Condition",
  "listingUrl": "https://www.ebay.com/itm/396512340987",
  "salePrice": 249.99,
  "currency": "USD",
  "shippingCost": 0,
  "shippingText": "Free delivery",
  "soldDate": "2026-08-09T00:00:00.000Z",
  "soldDateText": "Aug 9, 2026",
  "condition": "Pre-Owned",
  "buyingFormat": "Buy It Now",
  "bids": null,
  "itemLocation": "United States",
  "imageUrl": "https://i.ebayimg.com/images/g/AbCdEfGh/s-l500.jpg",
  "page": 1,
  "source": "ebay.com",
  "scrapedAt": "2026-08-10T12:41:07.882Z"
}

Auction rows look the same but carry "buyingFormat": "12 bids" and "bids": 12.

Measured field coverage

Measured on a real 300-row platform run on 2026-08-10: query nintendo switch, 2 pages, ebay.com. 300/300 rows real, 0 diagnostics, 54 s.

bids is only present on auction listings — 19% of that sample. No column was empty on every row. Median sale price in the sample: $126.75.

Input

• query — keywords. Leave it empty and you get one labelled, uncharged sample row.
• site — ebay.com, .ca, .co.uk, .de, .com.au, .fr, .it, .es, .ie.
• categoryId — eBay category ID (e.g. 625 for cameras). Verified to change results.
• condition — new, open box, refurbished, used, parts. eBay applies it server-side.
• minPrice / maxPrice — applied by eBay and re-checked on the parsed price.
• soldFrom / soldTo — date window, applied after parsing.
• currency — keep only listings in one currency.
• maxItems (default 120), maxPages (default 3, ~200 listings per page),

maxSessions (default 8), requestTimeoutSecs.

• proxyConfiguration — optional; leave it empty for the tuned default.

Every filter above was run live before being documented. Rows a filter drops are neither returned nor charged.

Input field names follow the conventions used across similar Actors, so switching is usually a drop-in change — query, maxItems and proxyConfiguration mean what you already expect them to mean.

A full example: everything sold in the last month, used or better, between $150 and $400, on the UK site.

{
  "query": "nintendo switch oled",
  "site": "ebay.co.uk",
  "condition": "used",
  "minPrice": 150,
  "maxPrice": 400,
  "soldFrom": "2026-07-15",
  "maxItems": 300,
  "maxPages": 3
}

The minimum viable input is just { "query": "nintendo switch oled" } — everything else has a working default.

Reliability

eBay refuses a large share of automated traffic outright, and it does not announce it: an interstitial arrives as HTTP 200 with a small body and zero listings. So this Actor judges the body content, never the status code — a 200 that contains no result cards is treated as a refusal, not as an empty search.

Sessions are warmed before they are used. A request fired straight at a search URL succeeds far less often than one made by a session that first visited the marketplace homepage and kept the cookies it was handed, then asked for the search page with a same-origin referer. Warmed sessions also paginate several pages deep without being challenged, where cold ones are stopped almost immediately. Those two numbers were measured, not assumed, on real platform runs.

A session that starts being refused is discarded and a fresh one is warmed on a new address — a burned eBay session stays burned, so retrying on the same one only wastes time. Each page gets up to maxSessions attempts. Egress is rotating datacenter addresses; no residential bandwidth is used, so there is no per-gigabyte surprise on the bill.

eBay also changes its result-card markup periodically, which silently turns a working parser into one that returns zero rows. The parser here accepts both the current and the previous card structure, and a page that genuinely yields nothing produces an uncharged diagnostic row rather than a silent empty dataset.

Billing

$0.70 per 1,000 sold listings ($0.0007 each), plus $0.001 each time a run starts. This is the cheapest eBay sold-listings scraper on the Apify Store.

Pay per event, event name listing. One event per unique real listing. Duplicates across pages are charged once. Never charged: the empty-input sample, filtered-out rows, EBAY_BLOCKED, NO_LISTINGS and NO_MATCHES diagnostics.

What a run actually costs:

Because charging happens per row after the row is pushed, a run that is blocked or matches nothing costs you the $0.001 start fee and nothing else. Filters are the cheap way to work: minPrice, condition and soldFrom drop rows before they are charged, so a narrow query is genuinely cheaper than a broad one you post-process yourself.

What this does not do

• No seller username or feedback score. eBay removed those from the search-result card; this actor

will not fabricate them. Use a seller-profile scraper if you need them.

• No per-item detail pages, so no item specifics, description HTML, or full photo set.
• No live/active listings — this is the sold and completed view only.
• No eBay API, no OAuth, no application key.
• No residential proxy.
• eBay caps a search at roughly 200 listings per page; very deep pagination eventually stops

returning new items and the actor writes an uncharged NO_LISTINGS row rather than looping.

• Non-US marketplaces publish fewer attributes on the result card, so condition and itemLocation

are sparser there than the 100% measured on ebay.com.

• Sold-listing coverage on eBay itself is roughly the last 90 days. soldFrom cannot reach further

back than eBay retains.

FAQ

How do I find what something sold for on eBay? Put the product name in query, set soldFrom to the start of the window you care about, and read salePrice and soldDate from each row.

How many sold listings per run? About 200 per page. maxPages up to 20, maxItems up to 2,000. A 300-row run took 54 seconds.

Do I need an eBay API key? No. Nothing to register, nothing to authorise.

Can I search a non-US eBay site? Yes — set site. Prices come back in that marketplace's currency and soldDateText in its date format; soldDate is normalised to ISO either way. Note that non-US sites publish fewer attributes, so condition is much sparser there than the 100% measured on ebay.com.

Does it include shipping in the price? No. salePrice is the item price and shippingCost is separate, so you can add them yourself or compare item-only prices.

How do I get a median or a price range out of this? Run the query, export the dataset as CSV or JSON, and aggregate salePrice yourself — the rows are one listing each, already parsed to numbers, so a median is a one-liner in a spreadsheet or pandas. Use currency to avoid mixing marketplaces in the same average.

What happens if eBay blocks it? Each page gets up to maxSessions freshly warmed sessions on new addresses. If all of them are refused you get one uncharged EBAY_BLOCKED row saying so, and zero listing charges.

Is the sample row charged? No. Empty input returns one _sample: true row and zero billable events.

Can I schedule it? Yes — a daily or weekly schedule on the same query gives you a rolling comp history. Each run is independent, so deduplicate on itemId if you merge datasets across runs.