eBay Scraper — Search, Listings, Prices, Items & Sellers
Pricing
from $3.00 / 1,000 listing rows
eBay Scraper — Search, Listings, Prices, Items & Sellers
eBay listings by keyword, search URL or item URL on 10 marketplaces incl. ebay.de. One row per listing: price, shipping, condition, Buy It Now or auction, bids, units sold, watchers, seller feedback, item specifics (item pages). Free filters: condition, buying format, price range, category, sort.
Pricing
from $3.00 / 1,000 listing rows
Rating
0.0
(0)
Developer
Tedj MEABIOU
Maintained by CommunityActor stats
0
Bookmarked
43
Total users
15
Monthly active users
9 hours ago
Last modified
Categories
Features
Share
An eBay scraper that turns active search results into clean rows: price, struck-through list price, shipping cost, condition, buying format, bids and time left, units sold, watchers, seller feedback score and percentage, item location and image. No API key, no login. It reads public search, seller and item pages on 15 eBay marketplaces, including eBay UK, Canada, Ireland, Austria and Switzerland, and bills you per listing row it actually delivers. Fields depend on what eBay displays. Use it for ebay price research, for bulk ebay product data, and as an ebay api alternative when the official Browse API's onboarding is more than your project needs.
Last verified working: 2026-11-19 — keyword, seller and Store searches, pasted search URLs, item URLs and onlyNew monitoring on ebay.com, multi-page searches on ebay.ca, and search pages on ebay.co.uk, .ie, .at, .ch and .de, read from exits in each marketplace's own country so prices and postage are the ones a local buyer sees. eBay's Akamai bot wall (since 2026-11-13) is passed the way a visitor passes it: each proxy session opens with one real-browser (Camoufox) page on its own IP, and the pages after it reuse that browser's cookies on the same IP. Since 1.8 a session that passed is also kept for your next runs, which reopen it on the same IP and need no browser at all (in a check on 2026-11-18 the first row came in 5.8 s instead of 30.5 s). Sold and completed listings still need a signed-in eBay account (re-checked with a real browser) and are refused for free, never answered with active listings.
The differentiator is the demand column. Most ebay listing scraper output tells you what sellers are asking. This one also tells you what shoppers are doing about it: how many units a listing has already sold, how many people are watching it, how many bids an auction has drawn and how long it has left, whether the seller takes offers, and how far the current price sits below the seller's own list price. Those five numbers are what separate a price you can copy from a price nobody pays.
Quick start
In Apify Console, no code. Type a keyword into Search keywords (the form starts with "mechanical keyboard"), press Start and open the Output tab: about 60 listings with price, shipping, condition, units sold, watchers and seller, for $0.24. Export gives CSV, Excel or JSON. For a file with listings only, set Status rows in the dataset to "None: listings only" before you start (statusRows: "none"); otherwise each search adds one free status row, and its market summary adds about 75 columns to a full CSV export.
From the API, one call, rows back:
curl -X POST "https://api.apify.com/v2/acts/kestrel~ebay-search-scraper/run-sync-get-dataset-items?token=$APIFY_TOKEN&format=csv" \-H 'content-type: application/json' \-d '{"queries": ["mechanical keyboard"], "maxResultsPerQuery": 20, "statusRows": "none"}'
That is 20 listings as CSV for $0.08. Drop &format=csv for JSON. Python, JavaScript and integration recipes are further down.
From an AI agent (Claude, ChatGPT, Cursor, any MCP client): add Apify's MCP server with this actor as a tool, https://mcp.apify.com?tools=kestrel/ebay-search-scraper (setup below), and ask in plain words, e.g. "What do used iPhone 13s sell for on eBay UK?". Every search's status row carries market.quick_take, the market in a few sentences, so the agent can answer without scanning every row.
How long a run takes. A first run in your account waits while a real browser gets past eBay's bot check: on 2026-11-19 the first page came 32 s to 2.6 minutes after the start (runs 2jnbl7CpEohVaixeq, GKsEigVOoOk1cw4m7, MjRnCVUCsld10DDfb, koWunted0cGQ6CzNq). Your later runs reopen the sessions that passed, and had 60 listings 4 to 6 s after the start (9AlfLmNXn9FrPRk81, RORZbbny5sKaPesYs). While a run works, its run card says how far it has got: listings so far, jobs done, pages read and how the bot check is going.
Pricing dates for API callers
The active price is $0.004 per delivered listing row (since 2026-10-27), $0.0035 on Apify's Scale plan and $0.003 on Business (since 2026-11-16). One page of 60 delivered listings costs $0.24 at the base rate. Set maxTotalChargeUsd separately in Apify run options.
中文:当前每条已交付商品记录为 $0.004,自 2026-10-27 起生效。60 条记录费用为 $0.24。请在 Apify 运行选项中设置费用上限。
eBay 商品搜索、价格、拍卖与卖家反馈:中文调用与语言设置
按关键词或 URL 获取商品、价格、运费、商品成色、竞价及来源展示的售出数量。它不是完整已售成交历史数据库。
domain 选择已支持市场,如 com、co.uk、de、fr、ca。sellers 可按卖家或店铺链接抓取;onlyNew 只交付新商品。每个市场使用其货币;没有 cn 市场或语言翻译开关。买家所在国家不会自动转换结果币种。
| 调用需求 | 参数与结果约定 |
|---|---|
| 小规模验证 | 使用下面 JSON;先检查结果和状态,再提高上限 |
| 语言与市场 | 按上面的真实字段选择,不要把中文说明当成自动翻译承诺 |
| 费用控制 | maxTotalChargeUsd 是 Apify 运行选项,不是输入 JSON 字段;按定价页的计费事件设置上限 |
| 数据完整性 | 检查状态行和 SUMMARY;空结果、来源错误与完整数据集是不同情况 |
{"queries": ["mechanical keyboard"],"pages": 1,"maxResultsPerQuery": 1,"sessions": 1}
eBay Angebote / enchères / subastas:固定市场与购买方式后比较价格。
This localized quick start uses the existing input contract. API users should send the JSON as actor input and set maxTotalChargeUsd separately in Apify run options. Source coverage and translation depend on the platform, not the language of this guide.
eBay scraper for AI agents: which input answers which request?
Use this actor when the request wants eBay's active listings for a keyword, a pasted eBay search URL or an item URL: asking prices, shipping, condition, Buy It Now or auction, bids, units sold, watchers and seller feedback where eBay displays them, and item specifics, quantity and category path from item pages. It also lists a seller's or an eBay Store's listings, and on a schedule it can deliver only new listings and price changes. It does not return sold or completed listings and sold prices (eBay shows those only to signed-in accounts; such a request ends free with an explanation), eBay Kleinanzeigen (a different site), or eBay Hong Kong and Philippines.
| user request | input JSON |
|---|---|
| listings for X | {"queries": ["mechanical keyboard"]} (first page, up to 60 listings, $0.24) |
| 200 listings for X | {"queries": ["mechanical keyboard"], "pages": 4, "maxResultsPerQuery": 200} |
| what the market for X looks like: price range, what sells, who sells | {"queries": ["mechanical keyboard"]}: every search's status row carries market (price and landed-price quartiles, median price by condition and format, units sold, top sellers, the 3 cheapest listings) and market.quick_take, the same in a few sentences, free |
| a clean CSV or sheet of the listings, nothing else | {"queries": ["mechanical keyboard"], "statusRows": "none"} (no status rows in the dataset; the run's SUMMARY and STATUS records keep them) |
| X in this eBay category (a category page you copied) | {"queries": ["iphone 15"], "category": "https://www.ebay.com/b/Cell-Phones-Smartphones/9355/bn_320094"} (an id such as "9355" works too) |
| 20 listings of X that at least 5 people watch | {"queries": ["mechanical keyboard"], "minWatchers": 5, "maxResultsPerQuery": 20, "pages": 5} (the cap counts listings that pass; listings under the floor are free) |
| 5,000 listings for X (more than eBay pages through for one search) | {"queries": ["iphone 15"], "maxResultsPerQuery": 5000} (cut into price bands that eBay shows in full) |
| the 20 cheapest X | {"queries": ["mechanical keyboard"], "sortBy": "price_low", "maxResultsPerQuery": 20} |
| new X under 100 | {"queries": ["mechanical keyboard"], "conditions": ["new"], "maxPrice": 100} |
| X without cases, covers or empty boxes | {"queries": ["iphone 13"], "excludeWords": ["case", "cover", "box only"]} (excluded listings are never billed) |
| exactly this model number, no near-matches | {"queries": ["Line 6 Spider IV 120"], "titleMustMatch": true} (titles without every word are left out free) |
| X with free shipping, located in the UK | {"queries": ["sony wh-1000xm5"], "domain": "co.uk", "freeShipping": true, "itemLocation": "domestic"} |
| carry on a big search that stopped at its time limit | the same input plus "excludeDatasetIds": ["<that run's defaultDatasetId>"] (what it delivered is skipped, free) |
| which vehicles this part fits | {"startUrls": ["https://www.ebay.com/itm/147556525326"], "allCompatibleVehicles": true} (every vehicle in compatibility: year, make, model, trim, engine; without the switch the first 20 and compatibility_count) |
| the seller's description as HTML | {"startUrls": ["https://www.ebay.com/itm/287719134364"], "includeDescriptionHtml": true} (description_html, next to the plain-text description) |
| auctions ending soon | {"queries": ["rolex submariner"], "buyingFormat": "auction", "sortBy": "ending_soonest"} |
| used or refurbished X on eBay Germany | {"queries": ["iphone 13"], "domain": "de", "conditions": ["used", "certified_refurbished", "seller_refurbished"]} |
| X inside one category | {"queries": ["gpu"], "category": "175672"} |
| this active-listing eBay search | {"startUrls": ["https://www.ebay.com/sch/i.html?_nkw=gpu&LH_BIN=1"]} |
| X on eBay UK (or Canada, Ireland, Austria, Switzerland) | {"queries": ["iphone 13"], "domain": "co.uk"} (ca, ie, at, ch) |
| everything one seller has listed | {"sellers": ["vipoutlet"], "pages": 3} |
| X from these sellers only | {"queries": ["iphone"], "sellers": ["verizon", "vipoutlet"]} |
| this eBay Store's listings | {"startUrls": ["https://www.ebay.com/str/mytrendytrunk"]} |
| details of one item | {"startUrls": ["https://www.ebay.com/itm/287719134364"]} (one item-page row, $0.004) |
| search results with item specifics, description, variations and identifiers | {"queries": ["lego 10294"], "includeDetails": true} (add-on, $0.0025 per enriched listing on top of $0.004; active from 2026-12-02). Until then: pass the item URLs into startUrls in a second run ($0.004 per item row) |
| sold/completed listings or sold prices | Not available: eBay shows them only to signed-in accounts, so the run ends free with an explanation. units_sold on active rows is the public demand signal |
| only the new listings, every morning | {"queries": ["mechanical keyboard"], "sortBy": "newly_listed", "onlyNew": true} on a schedule: the first run delivers the page, later runs only listings it has not delivered before |
| new listings and price drops of a search | {"queries": ["rtx 4090"], "onlyNew": true, "includePriceChanges": true} on a schedule: rows carry change = new or price_change and previous_price |
| what buyers say about seller X | {"feedbackSellers": ["vipoutlet"], "feedbackRating": "negative"} (one feedback row per entry: rating, comment, item; add-on, active once its price is set) |
| what the auctions I watch sold for | {"queries": ["pokemon psa 10"], "buyingFormat": "auction", "sortBy": "ending_soonest", "includeSoldAuctions": true} on a schedule: each auction is read again after it ends and a sold one comes as change = sold with its final price (add-on, active once its price is set) |
Output rows share one dataset and are told apart by type. An item row (billed, $0.004 each) has fields item_id, title, url, marketplace, price, currency, list_price, discount_pct, shipping_cost, free_shipping, condition, buying_format, bids, time_left, best_offer, units_sold, watchers, returns, seller, seller_feedback_pct, seller_feedback_count, location, image, results_total, coupon, coupon_price; values depend on the source display. Item-URL rows also fill quantity_available, breadcrumbs, category, item_specifics and seller_url, plus the return policy (returns_accepted, return_days, return_shipping_paid_by), seller_type, the UK/FR/IT Buyer Protection fee split and parts compatibility (compatibility_count, compatibility). A status row (free, one per keyword or URL) says ok, no_results or error, with items, filtered, duplicates, results_total, left_out (what was left out free, by reason), a note and market, a free summary of the listings the search delivered. A feedback row (the seller-feedback add-on) is one entry of a seller's feedback profile. A search that matches nothing is a free no_results row; an eBay refusal is error, never an empty success, and its note says what to do; bad input ends the run with the reason on the run card and nothing charged. statusRows: "none" keeps status rows out of the dataset (listings only, for exports), "problems" keeps only the ones with something to say.
eBay listings: what does this scraper collect?
Every item row uses the same columns; fields not present in the source can be null. Item-page fields are filled on item-URL rows.
| Field | What it is |
|---|---|
item_id | eBay's numeric listing id — the number in any /itm/ URL |
title | Listing title as the seller wrote it |
url | Canonical item page URL |
query / search_url | The keyword, and the exact URL the page was read from |
page / position | Where the listing sat in eBay's own order, counted across pages |
domain / marketplace | Supported host such as com, de or fr, and the country name |
price / price_display / currency | The number, the string exactly as printed, and the currency read from the price itself. price is null on the occasional MAP-restricted listing where eBay prints "See price" instead of a number — price_display then says so, and price bounds drop those rows free |
list_price / list_price_display / discount_pct | The struck-through original and how far below it the listing sits |
shipping_cost / shipping_display / free_shipping | Delivery charge as a number, the full line, and the free-delivery flag |
condition | Brand New, Pre-Owned, Open Box, Certified - Refurbished, For parts … |
buying_format | buy_it_now or auction |
bids / time_left | Auction rows: bid count and eBay's own countdown text |
best_offer | Whether the seller accepts offers |
units_sold | How many units this listing has sold |
watchers | How many shoppers are watching it |
returns | The returns line, e.g. "Free returns" or "30 days returns" |
seller / seller_feedback_pct / seller_feedback_count | Username or store, positive-feedback percentage, and the score expanded from eBay's 40.6K shorthand to 40600 |
location | Where the item ships from |
image | Listing image URL |
rating / ratings_count | Product star rating, when the listing is matched to a catalogue product |
results_total | How many listings eBay reports for the whole search |
coupon / coupon_price | The coupon as eBay prints it ("From £213.39 with coupon", "Ab EUR 119,95 mit Gutschein", "10% off with code PARTSAVE10 · Ends 26 Nov • Max $75.00") and the price with it; a multi-buy offer ("$5 off every 3 items with coupon") has no coupon_price |
quantity_available, breadcrumbs, category, item_specifics, seller_url | Item-page fields (item-URL rows): availability text, category path, leaf category, seller's specifics table and store URL |
details_fetched, details_error | Item-details add-on and item-URL rows: whether the item page was read and merged in, and why not |
description, images, variations | The seller's description as text, the full-size gallery, and each variation with its attributes, price and stock |
brand, model, mpn, gtin, upc, ean, isbn, epid | Identifiers from the page's product data and item specifics (epid = eBay catalogue product id) |
category_id, listing_started, quantity_sold | Leaf category id, when the listing started, units sold on the item page |
seller_username, seller_feedback_score, seller_items_sold, seller_joined, seller_store_url | The seller's username (Store names differ), feedback score, items sold, member-since and Store URL |
returns_accepted, return_days, return_shipping_paid_by | Item pages: whether returns are accepted, the return window in days, and who pays return shipping (seller or buyer), from the page's product data or its returns line in the site's language |
seller_type, buyer_protection_fee, price_before_fee | Item pages: private or business where eBay prints it (ebay.co.uk and other EU sites). A private seller's price on ebay.co.uk, .fr and .it includes eBay's Buyer Protection fee; these two columns split it out (£120.30 = £115.00 + £5.30) |
compatibility_count, compatibility | Item pages of parts: how many vehicles eBay says the part fits, and the first 20 of them (year, make, model, trim, engine, notes) |
coupon_code | Item pages: the coupon's code, e.g. PARTSAVE10 |
change, previous_price | onlyNew runs: new, price_change or unchanged (with emitUnchanged), and the price this search last delivered |
description_html | With includeDescriptionHtml, item pages: the seller's own description markup (up to 200,000 characters) |
compatibility_complete | With allCompatibleVehicles, item pages of parts: whether compatibility holds every vehicle eBay counts |
fetched_at | UTC timestamp |
A second row type, status, is emitted once per keyword or URL and is always free: kind, target, status (ok / no_results / error), items, filtered, duplicates, pages, results_total, error, left_out, note, market, fetched_at. Read it to know that a job ran and what it produced, without inferring anything from row counts. With statusRows: "none" (or "problems") they stay out of the dataset, or only the ones with something to say go in, and the run's STATUS record lists every one. left_out counts the listings left out free, by reason (title, fewer_words, price, excluded_words, free_shipping, location, watchers, units_sold, delivered_before …); note says in words what else happened (an empty answer read again on another IP, or how to carry on a walk the run timeout stopped). The same totals are in the run's SUMMARY record and on the run card.
eBay search results by keyword: how do I get them?
To scrape eBay search results by keyword, give it keywords and a depth. Each page carries 60 listings.
{"queries": ["mechanical keyboard", "keycap set"],"domain": "com","pages": 2,"maxResultsPerQuery": 60}
That run reads two pages per keyword, keeps the first 60 listings of each in eBay's own order, and charges for 120 rows at most — fewer if a keyword returns fewer.
How do I get more than eBay's 2,000 listings for one search?
eBay pages through about 2,000 listings of any one search (34 pages of 60), however many it says it found. Ask for more and the actor cuts the search into price bands that eBay shows in full: set maxResultsPerQuery above 2,040 (up to 100,000), or maxItems above 2,040 with one keyword, and leave pages alone. A band whose first page reports more than 1,900 listings is cut in two, and that first page is only a probe: its listings come again in the halves and are billed once, there. Bands run from the cheapest up (from the dearest down for sortBy: "price_high"), so a price sort stays in order; position then counts across the whole search and results_total is the whole search's count. Listings that share one exact price beyond eBay's 2,000 cannot be told apart by any price band, so those stay at eBay's limit. Monitoring runs (onlyNew) read the pages they are given and do not walk bands. Other scrapers' splitByPrice: true asks for every band.
Since 1.10 a walk delivers and bills page by page, so the run timeout, the spending limit or a page eBay refuses keeps every listing already read (test_a_band_walk_delivers_page_by_page_and_keeps_what_it_read_when_a_page_is_refused). Since 1.11 a walk reads eBay's full view: on some searches eBay shows a "streamlined" view ("We've streamlined your search results to show you the best listings") that leaves listings out, and one $640-660 "iphone 15" band had 249 results that way and 367 in the full view (eBay's own "See more results" link, _blrs=recall_filtering). At about 390 listings a minute, a walk of more than ~20,000 listings needs more than the default one-hour timeout: raise the timeout, or let it stop and carry on. A walk the timeout stopped says so on its status row and run card, with the exact value to paste: run the same input again with "excludeDatasetIds": ["<the first run's defaultDatasetId>"], and the listings already delivered are skipped free, never billed twice. On 2026-11-18 a 2,500-listing walk stopped at its 300 s timeout with 1,380 delivered and the value on its run card (l6SLUJnmhzRUhZj2B); the same input with that value delivered the other 1,122 and left the 1,378 it met again out, free (Rq7Mo0kmUUpBzAi2e). If the platform restarts a run (moves it to another server, reboots or resurrects it), the run reads its own dataset back and delivers nothing twice.
{"queries": ["iphone 15"],"maxResultsPerQuery": 5000}
How do I use my own eBay search URL?
Build an active-listing search on a supported eBay marketplace and copy its URL into startUrls. Its search filters are retained; paging is controlled by the Actor. A sold/completed search URL (LH_Sold=1 or LH_Complete=1) is refused before anything is fetched or charged, because eBay answers it only for signed-in accounts; it is never quietly turned into an active-listing search. The marketplace comes from the URL, so one run can mix supported countries.
{"startUrls": ["https://www.ebay.com/sch/i.html?_nkw=lego+star+wars&_sacat=19006","https://www.ebay.de/sch/i.html?_nkw=nintendo+switch+oled"],"pages": 3}
Item URLs (https://www.ebay.com/itm/<id>, any supported marketplace) give one item-page row each, with the item specifics table, quantity, category path, seller store, return policy, seller type and parts compatibility added to the usual columns, at the same $0.004. An item URL whose listing has ended (sold, or closed unsold) is a free no_results status row that says so and when it ended; it is never billed as an active listing. Item pages are read with eBay's nordt=true, because without it an ended catalogue listing is answered with its product page and another seller's live offer.
How do I search inside an eBay category?
Give category the category's numeric id ("9355"), or paste the category's URL as you copied it: a category page (https://www.ebay.com/b/Cell-Phones-Smartphones/9355/bn_320094), a category search (/sch/9355/i.html) or any search URL with _sacat=9355. Every keyword of the run is then searched inside that category. A URL also sets the marketplace when you left domain at its default, because category ids differ between eBay sites; a domain you set wins. A category URL that names no category (eBay's curated pages such as Brand Outlet) ends the run free with the reason (test_a_category_url_is_read_for_its_id_and_marketplace, test_a_category_that_names_no_category_ends_free). memo23's motorsCategory preset and caffein.dev's categoryId / subcategoryId are read too.
A category on its own, without a keyword, and a /b/ category page pasted into startUrls, are tried as eBay's search. On 2026-11-19 eBay answered both with its category browse page instead: a merchandised view (about 45 catalogue cards a page, the same listing on several pages), not the category's listings; on ebay.de a keyword-less category search was redirected to such a page. Before 1.12 such a URL ended as a silent "found nothing" (run p1J2jwBYd3xnzcjow on 1.11.1); now the job's status row says what eBay answered and how to get the listings (add a keyword), and nothing is charged (run wHHpICkGdVnQV1VBu; test_a_category_page_or_a_category_alone_is_never_a_false_no_results, test_a_redirect_to_a_category_page_is_told_apart_by_its_canonical_link). Where eBay does answer a category with search results, they are delivered as usual. With a keyword, a pasted category page works as a filter: "iphone 15" in https://www.ebay.com/b/Cell-Phones-Smartphones/9355/bn_320094 gave 20 listings from _sacat=9355 (run V6semgNiRUGdenu9S).
How do I leave out listings with a word in the title?
Set excludeWords, e.g. ["case", "cover", "box only"]. Single words are sent to eBay as minus terms (iphone 13 -case -cover), so eBay leaves those listings out before it sends the page and they cost nothing. Phrases, and anything eBay's own matching lets through, are dropped here before billing, counted in the status row's filtered and in SUMMARY as excluded. Matching is on whole words: psa drops "PSA 10 Charizard" but not "PSA10" or "capsule". You can also type eBay's own syntax into a keyword ("iphone 13 -case"). On a seller's listings without a keyword, only the title filter applies. Other scrapers' excludeKeywords and negativeKeywords work too, as a list or as one string of words.
How do I get only listings whose title has every word?
Set titleMustMatch: true. eBay's search matches loosely: for "Line 6 Spider IV 120" it also returns Spider IV 15, Spider V 120 and FBV cables, and other scrapers bill every one of them (on 2026-11-18 such a search billed 20 rows here before 1.10, none with every word: run njtQYO6xWW9fiAwvn; with titleMustMatch the same input billed 0 and said "60 left out free: 60 without every search word in the title": run w0YUVoeJJAPCoX2ah). With titleMustMatch, a listing whose title lacks one of the keyword's words is left out before billing, free, and counted in left_out as title. Case, accents and punctuation do not count, so 18V-55, 18v 55 and 18v55 are the same; a quoted phrase must appear in that order; -word and (a,b) follow eBay's syntax; 120 does not match 1200. maxResultsPerQuery then counts matching listings, within your pages. For a pasted search URL its own keyword (_nkw) is used; seller-only and item jobs have no keyword and are not filtered. memo23's titleMustMatchQuery works too, and its titleMustInclude (words of your own, checked on every job).
eBay's "Results matching fewer words" padding is never delivered, with or without this setting: on a page with a few exact matches, the cards under that heading are left out free (fewer_words), never count toward a cap, and the page they start on is the last one read; a page with no match delivers nothing. "iphone 15" between $0.01 and $0.99 had 1 match and 8 iPhone 7 Plus cards under that heading on 2026-11-18, and 1.9 billed all 9 (test_only_ebays_fewer_words_notice_cuts_a_page, test_ebays_fewer_words_cards_are_never_billed). eBay's other notice, "We've streamlined your search results to show you the best listings", stands above real results and cuts nothing.
How does the price filter save me money?
minPrice and maxPrice become eBay's own _udlo and _udhi parameters, so eBay does the filtering before it sends the page. You pay for the pages that come back, not for thousands of rows you throw away afterwards. The bound is then checked a second time here, because eBay applies its own bound to price plus shipping in some views and a row can still arrive outside your range — anything that fails the second check is dropped and never billed. Those drops show up in the filtered count on the status row.
{"queries": ["vintage pocket watch"],"minPrice": 50,"maxPrice": 400,"buyingFormat": "auction","sortBy": "ending_soonest","pages": 2}
buyingFormat, conditions, category, sortBy, freeShipping (eBay's LH_FS), returnsAccepted (LH_RPA), itemLocation (domestic = eBay's "US Only" / "UK Only", worldwide) and searchInDescription (eBay's "Search including description", LH_TitleDesc) are server-side in exactly the same way. Free shipping and the location are checked again on each card before billing, because eBay lets a few listings from abroad through its own location filter. conditions is a multi-select because eBay grades refurbished in four separate tiers — Certified, Excellent, Very good, Good — and a filter that sends only "refurbished" silently drops most of that market.
Caps count the listings you get, not the listings read
With a filter that is checked on each card (price bounds, excludeWords phrases, freeShipping, itemLocation, other scrapers' minWatchers, minQuantitySold, sellerCountry, model years), titleMustMatch or onlyNew, maxResultsPerQuery counts the listings that pass: the pages you allow are read one at a time until it is met. Before 1.12 the cap counted listings read, so 20 asked with a watcher floor gave 11 from a page that held 16 (test_a_card_filter_cap_counts_the_listings_delivered_not_read). On 2026-11-19 memo23's own shape, minWatchers: 3 with maxItems: 30, read 2 pages and delivered 30, with 77 listings under the floor left out free (run MjRnCVUCsld10DDfb). Another scraper's per-search count with such a filter (memo23's maxItems with minWatchers or titleMustMatchQuery, caffein.dev's count with sellerCountry) is read the way those scrapers mean it: pages are read on until that many listings pass, up to eBay's 34-page stop (test_another_scrapers_count_with_a_card_filter_reads_on_until_it_is_met). A monitor still reads only the pages it is given, since deeper listings are old ones it never saw. Listings left out are never billed.
Can I add item specifics to eBay search results?
Yes. Set includeDetails: true (the item-details add-on, off by default, active from 2026-12-02). Every delivered search listing is then enriched with its item page: the seller's description as text, the full image gallery, variations (each with its attributes, price and stock), identifiers (brand, model, MPN, GTIN/UPC/EAN/ISBN, eBay product id), the item specifics table, quantity and quantity sold, category path and leaf category id, and the seller's username, feedback score, items sold, member-since date and Store URL. The search card's own columns stay as they were. It costs $0.0025 per enriched listing ($0.0022 on Scale, $0.002 on Business) on top of the $0.004 listing row, and only when the page was read: a page eBay refused leaves a plain listing row (details_fetched: false, details_error says why) that is not charged for details. The spending limit pays rows first; no item page is read for details it cannot pay for. Before the add-on is active a run with includeDetails returns plain rows, charges nothing for details and says so in its status message.
Item URLs in startUrls always carry these fields, at the listing price. Two free switches add to item-page rows: includeDescriptionHtml keeps the seller's description markup in description_html (from the description page already read; test_description_html_is_on_item_page_rows_when_asked_and_one_shape_holds), and allCompatibleVehicles lists every vehicle a part fits in compatibility, read 20 at a time through the item page's own compatibility-table paging (the same request the page sends when you click its "next page"), up to 2,000, with compatibility_complete (test_every_compatible_vehicle_through_the_tables_own_paging). A brake-pad listing gave all 124 of its 124 vehicles and its 28,176-character description markup in 16 s (run QkqYJWcFGP5Phh47m).
Without the add-on
Two runs do it too. Search-card rows do not carry item_specifics, quantity_available, category or seller_url; item pages do. Run the search, pick the listings you care about, and pass their urls into startUrls in a second run. Keep the two runs separate: within one run an item id is delivered only once, so a search row and an item URL for the same listing would keep only the first.
Billing is $0.004 per delivered row in each run: ten search rows cost $0.04, and ten item-page rows for them another $0.04. A refused item page is an error status row and costs nothing.
eBay price research: how do I compare a market?
Point the actor at a product's keywords, take the whole first page or two, and the distribution answers the question. This is bulk ebay data extraction with the analysis columns already attached. price plus shipping_cost is the real landed cost, and it is the number ebay buy it now shoppers actually compare. units_sold tells you which of those prices has ever transacted. The ebay units sold figure is the strongest signal on the page. The ebay watchers count tells you where interest sits without a sale yet, and a listing with many ebay watchers and no sales is usually priced above the market. results_total tells you how crowded the listing is. condition lets you split the used market from the new one, best_offer shows which sellers will negotiate — an ebay best offer flag on most of a page means the printed prices are ceilings, not clearing prices — and ebay seller feedback lets you discount the outliers from accounts nobody buys from.
A practical routine for resellers and sourcing teams:
- Run the keyword with
sortBy: "price_low"andpages: 2to find the floor. - Run it again with
buyingFormat: "auction"andsortBy: "ending_soonest"to see what is clearing right now. That is ebay auction data in its most useful form:bidsandtime_lefton those rows are live competition, and a lot of bids with little time left is a price about to be proved. - Sort your export by
units_solddescending. The top rows are the listings that have proven the price. - Compare
priceagainstlist_price— a largediscount_pctacross many sellers is a category in decline, not a bargain. - Schedule the same run daily and diff on
item_idto watch new listings, price moves andunits_soldgrowth.
This is the ebay market research loop that ebay reseller tools charge a monthly subscription for, except it lands in your own dataset and you pay by the row.
The free market summary of every search
Since 1.12 every search's status row carries market, and SUMMARY.markets the same per search, free: a digest of the listings that search delivered, so an AI agent or a sheet reads the market without scanning every row.
priceandlanded_price(price plus shipping):min,p25,median,p75,max,mean, in the currency most rows carry (other_currencycounts the rest,no_pricethe "See price" listings);by_conditionandby_format: listings and median price per condition (Brand New, Pre-Owned …) and per buying format;units_sold,listings_with_salesandmedian_price_with_sales: what has already sold, and at what price;watchers,auction_share,best_offer_share,free_shipping_share;top_sellers: the five sellers with the most listings, with their units sold and feedback;cheapest: the three cheapest listings by landed price, with links;quick_take: the same market in a few plain sentences, for a person or an AI agent that reads one line. For "mechanical keyboard" on 2026-11-19 (run ZyNkPks78nd2FcDsT): "60 listings: median $49.99, middle half $42.59–$76.81 (with shipping: median $51.12). 17 of 60 have sold (840 units), at a median $49.99. By condition: Brand New $46.00 (26), Certified - Refurbished $56.77 (18), Pre-Owned $39.99 (7). 32% accept offers, 75% free shipping. Most listings: corsairdirect (10, 279 sold). Cheapest with shipping: $12.99." (test_the_market_summary_says_the_market_in_plain_sentences).
It covers the listings delivered (after your filters and caps), not eBay's sold history. Turn it off with marketSummary: false; memo23's includeSummary works too (test_every_search_status_row_carries_a_free_market_summary, test_the_market_summary_of_a_search_page). For "mechanical keyboard" on 2026-11-19 (run zNOfisgvq1Oz1Auac): 60 listings, median $50.63 with quartiles $44.99 to $99.77, 836 units already sold, corsairdirect the top seller with 9 listings and 279 sold, the cheapest landed at $12.24.
eBay seller data: how do I find who dominates a category?
The sellers view groups the same rows by who is selling. seller_feedback_count is expanded from eBay's abbreviated card display, so 40.6K becomes 40600 and you can sort on it numerically. Combined with units_sold, location and condition, this ebay seller feedback view is enough to find the handful of accounts that dominate a category, to check whether they are stores or private sellers, and to see how their pricing differs from the long tail. Auction cards do not print the seller line, so those search rows can lack seller data; an item-URL row for the listing has it.
eBay marketplaces: which ones are supported?
Fifteen: com, co.uk, de, fr, it, es, ca, com.au, at, ch, ie, nl, pl, com.sg, com.my. Set domain for keyword and seller searches ("ebay.co.uk", "UK" and "EBAY_GB" work too), or paste marketplace URLs into startUrls and mix as many as you like in one run.
eBay UK, Ireland, Canada, Switzerland and Austria joined in 1.6. Since August 2026 they had answered every plain HTTP request with eBay's /splashui/challenge page, a JavaScript check that a real browser clears by itself. On 2026-11-17 the run's browser passed on each of them, and the pages after it (search page 2 and on, item pages) read normally on the same IP. Hong Kong and the Philippines are recognised in pasted URLs (prices and labels come out right) but not offered. eBay Belgium is benl./befr.ebay.be, not www.ebay.be.
Condition labels, buying-format wording and the shipping line are parsed in each site's own language, so a German or Austrian run returns Gebraucht and Sofort-Kaufen rather than nulls. currency is read from the rendered price rather than assumed from the marketplace, because eBay localises the price block to the caller's egress.
eBay prices every page for the visitor's own country, so where the run reads from matters. Since 1.10 every exit of a run is in the marketplace's own country (US for ebay.com, GB for ebay.co.uk …): the datacentre pool is pinned to that country where Apify has IPs there, and otherwise the run starts on residential exits in that country. Before 1.10 only the residential exits were pinned, and runs started on datacentre exits that are mostly in the US: on 2026-11-18 a US exit read ebay.co.uk with 14 of 19 private sellers' prices £0.47-0.98 low (eBay's Buyer Protection fee without UK VAT; runs wkKkn1jALejlqHFpA against 7ASD7hmpUWvAlEZr1), and ebay.com.au with 3 of 17 rows without postage (Lt7ZWYTT7Dnx0YCaN against PZ3vJKHjVHhFGYFnA). The same inputs now gave 16 of 16 prices equal to the UK view (48bVEoqE77K9UwaoQ) and 17 of 17 equal to the Australian one, none without postage (ycrhSP9PYVY7BGAfN). Every page also names the postal code eBay priced it for; a page priced for a buyer abroad is read again from the marketplace's country (it happens on residential exits too: a US-pinned exit was priced by eBay for Vancouver, V6B2A3, and read again from the US in run GGkec6UAW8jQvvL7V), and if nothing can fix it (an account without residential proxies) the run says so in SUMMARY.buyer_view and on the run card (test_a_page_priced_for_a_buyer_abroad_moves_the_run_to_residential_in_the_country). US, German, French, Italian and Spanish postal codes share one shape, so on those sites the pinning, not the page check, is the guard.
On ebay.co.uk (and, by the same rule, .fr and .it) a private seller's price includes eBay's Buyer Protection fee: £0.10 plus 7% of the first £20, 4% up to £300 and 2% up to £4,000. On 2026-11-18 a private listing showed £120.30 on its search card and on its item page alike, which is £115.00 plus a £5.30 fee; a business seller's £159.00 carried no fee. price is always the price as eBay shows it, the same on search and item rows. Search cards do not say which sellers are private, so the split is on item-page rows (item URLs and the item-details add-on): seller_type, and for private sellers buyer_protection_fee and price_before_fee. The split is made only on a page priced for a UK buyer: on a price eBay made for a buyer abroad it would be wrong (test_no_fee_split_on_a_price_eBay_made_for_a_buyer_abroad).
How does it compare with the other eBay scrapers?
The five eBay scrapers on Apify with the most users in the last 30 days, read from their own input forms and READMEs on 2026-11-19 (caffein.dev's is sold listings only). ✓ for this actor means a named offline test or a dated run shows it.
| Feature | This actor | memo23 | automation-lab | delicious_zebu (listing, details) | caffein.dev |
|---|---|---|---|---|---|
| Active listings by keyword, search URL, seller or Store | ✓ | ✓ | ✓ | ✓ | – |
| Sold and completed search | ✗ eBay asks for a sign-in; such a request ends free | ✓ (eBay's app API) | ✗ | ✗ | ✓ |
| What a monitored auction sold for | add-on, once its price is set | ✗ | ✗ | ✗ | ✗ |
| No API token that can expire | ✓ | ✗ | ✓ | ✓ | ✓ |
| Item specifics, gallery, variations, identifiers | ✓ item URLs; search rows via the add-on from 2026-12-02 | ✓ | ✗ | ✓ (details) | ✗ |
| Description as HTML | ✓ includeDescriptionHtml | ✗ (text) | ✗ | ✓ (details) | ✗ |
| Every vehicle a part fits | ✓ allCompatibleVehicles | ✗ | ✗ | ✗ | ✗ |
| Market summary per search | ✓ free, every search | sold searches, a paid row | ✗ | ✗ | ✗ |
| Category by id or pasted category URL | ✓ | ✓ | id | ✓ | id |
| A whole category without a keyword | ✗ eBay's site answers with its browse page; said, free | ✓ | ✗ | ✓ | ✗ |
| Past eBay's ~2,000 per search | ✓ price bands | – | ✗ | ✓ | – |
| Only new listings and price changes on a schedule | ✓ | ✓ | ✗ | ✗ | ✗ |
| Watcher, sold and title filters before billing, caps counting what you get | ✓ | ✓ | ✗ | ✗ | – |
| Search in descriptions too | ✓ searchInDescription | ✗ | ✗ | ✓ | ✗ |
| Seller feedback (reviews of a seller) | add-on, once its price is set | ✓ | ✗ | ✗ | ✗ |
| Discord or Telegram alerts | ✗ (Apify's own integrations and webhooks) | ✓ | ✗ | ✗ | ✗ |
| Marketplaces | 15 | 17 | 8 | 8 | – |
| Other scrapers' inputs run unchanged | ✓ | ✗ | ✗ | ✗ | ✗ |
Proof for this actor, by row: test_a_keyword_delivers_rows_and_charges_exactly_them; test_a_sold_request_ends_free_with_the_reason_and_fetches_nothing; test_sold_auctions_are_read_once_after_they_end_and_billed_as_sold_items; reads public pages only; test_item_urls_carry_the_details_without_the_add_on_charge; test_description_html_is_on_item_page_rows_when_asked_and_one_shape_holds; test_every_compatible_vehicle_through_the_tables_own_paging; test_every_search_status_row_carries_a_free_market_summary; test_a_category_url_is_read_for_its_id_and_marketplace; test_a_category_page_or_a_category_alone_is_never_a_false_no_results; test_a_deep_walk_returns_every_listing_once_past_ebays_stop; test_only_new_delivers_everything_once_then_only_what_is_new; test_a_card_filter_cap_counts_the_listings_delivered_not_read; test_search_in_descriptions_and_other_scrapers_names_reach_ebay; test_seller_feedback_rows_are_billed_one_event_each_up_to_the_cap; test_the_new_marketplaces_are_offered; test_rival_shaped_inputs_deliver_paid_rows. Live runs of 2026-11-19: market summary zNOfisgvq1Oz1Auac, every vehicle and the description markup QkqYJWcFGP5Phh47m, a category URL V6semgNiRUGdenu9S, a /b/ page said free wHHpICkGdVnQV1VBu, caps counting what you get MjRnCVUCsld10DDfb, memo23's and delicious_zebu's names hcfqYsOmaUSLI8pde.
Where other scrapers fail, and what this actor does instead
From the issues and reviews of the other eBay scrapers on Apify (read 2026-11-18). Each claim on the right is checked by the named test in this actor's offline test suite, or by the dated check.
| Where other scrapers fail (their users' reports) | What this actor does | Proof |
|---|---|---|
| Blocked by eBay: "Starting today almost all have 403 status" (Nov 10), "No results being returned" (Nov 12, open), four eBay scrapers under maintenance | Each proxy session opens with one real-browser page on its own IP; since 1.8 the sessions that passed are reopened by your next runs on the same IP, with no browser at all. Cookies are reused only on the IP that earned them, checked before every reuse | test_one_browser_solve_per_session_and_host_then_curl_replays_its_cookies, test_a_cached_session_reopens_on_its_own_ip_and_needs_no_browser, test_a_cached_session_now_on_another_ip_never_replays_its_cookies; the same input 10 times in a row on 2026-11-18: 1.7.2 needed 2-7 browser pages every run (first row after 27-122 s), 1.8.2 one browser page in ten runs (first row after 3-13 s from the third run on; runs NMdbhqwjcQlcaSgfn, s1cShH8IAmJ3YT9yB) |
| "Only getting 45 results with a limit of 200", "Larger scrapes only get a fraction of the results" | Past eBay's ~2,000 per search, the search is cut into price bands eBay shows in full | test_a_deep_walk_returns_every_listing_once_past_ebays_stop, test_more_listings_than_ebay_pages_through_come_in_price_bands |
| "Run FAILED: eBay API 401, bearer token expired" (Nov 13) | Reads the public web pages: no API token to expire | – |
| "Completed listings ('Ended' are in the results as well)"; "Returned $17k for $3.5k sale" | A sold or completed search ends free with the reason, never answered with active listings; an ended item URL is a free no_results row; the auctions you monitor come back with their own final price, read off their own page (add-on) | test_a_sold_request_ends_free_with_the_reason_and_fetches_nothing, test_an_ended_item_url_is_reported_free_never_billed_as_an_active_listing, test_sold_auctions_are_read_once_after_they_end_and_billed_as_sold_items |
| "UK sold prices: Buyer Protection VAT and missing postage", "eBay served us the page for a non-UK buyer" | Every exit is in the marketplace's own country and every page's ship-to code is checked, so UK prices carry the VAT on eBay's fee and AU rows keep their postage; item-page rows split the Buyer Protection fee out of a private seller's price, only on a UK-priced page | test_a_page_priced_for_a_buyer_abroad_moves_the_run_to_residential_in_the_country, test_non_us_marketplaces_pin_datacentre_exits_to_their_country_and_fall_back_to_residential, test_uk_private_seller_prices_include_the_buyer_protection_fee; re-checked 2026-11-19 by the daily ebay.co.uk check: read from exits in GB, 0 pages priced for a buyer abroad (run 1G925IjjY0Tf7w0cn) |
| "Most returned rows are near-matches, which we discard but pay for" | titleMustMatch: titles without every word of the keyword are left out free | test_title_must_match_never_bills_a_near_match (memo23's own input, which billed 20 near-matches on 1.9.2: run njtQYO6xWW9fiAwvn; re-checked 2026-11-19: 2 listings with every word delivered, 57 "fewer words" cards and 3 near-matches left out free, run KFWY1oduz0uf99KSs) |
| "A 25-store batch fired back-to-back failed 25/25" (a store monitor, Nov 14), and its developer's "my honest recommendation is to migrate your workflow" | That monitor's own input runs here: incrementalMode, stateKey and emitUnchanged are read, so a daily store monitor bills only what is new | test_blackfalcondata_incremental_monitor_bills_only_what_is_new, test_a_state_key_names_the_memory_and_emit_unchanged_delivers_the_rest |
| "A silent empty result is indistinguishable from no sales" | An empty first page of a seller or a monitored search is read again on another IP before it counts as nothing found, the run card counts the searches that found nothing, and a search eBay has no listing for says what to try | test_an_empty_seller_page_is_read_again_before_it_counts_as_nothing (1.9.2 said "0 listings" for a seller with 96: run cptr3Uw0dzibxXBtw), test_a_search_that_finds_nothing_says_what_to_try |
| "You'd pay a second time for the 11,416 items you already have" | A big walk delivers page by page and keeps what it read; excludeDatasetIds carries it on without billing anything twice; a restarted run never re-bills | test_a_band_walk_cut_by_the_run_timeout_says_how_to_carry_on, test_exclude_dataset_ids_carry_on_without_paying_twice, test_a_restarted_run_never_delivers_or_bills_twice |
| "How to exclude graded cards from results?" | excludeWords: single words sent to eBay as -word, phrases dropped before billing | test_exclude_words_reach_ebay_as_minus_terms_and_filter_titles_free, test_other_scrapers_exclude_spellings_are_read_with_either_type |
| "Compatibility" (parts), "Return policy and coupons" requests | Free on item-page rows: the vehicles a part fits, the return policy, the coupon and its code; search rows carry the price with a coupon | test_parts_compatibility_first_table_page_and_the_count, test_return_policy_from_schema_or_from_the_returns_line, test_item_page_coupon_code_terms_and_price, test_card_coupon_lines_give_the_coupon_price_and_keep_offers_as_text |
| "The Actor returned Succeeded with 0 items" | An eBay refusal is an error row, never an empty success; a run where everything was refused fails | test_a_refusal_is_reported_as_an_error_not_as_no_results, test_every_job_failing_fails_the_run |
| "Results matching fewer words" padding billed as matches | eBay's zero-result page delivers nothing, and on a page with a few exact matches the cards under eBay's "fewer words" heading are left out free | test_a_zero_result_page_never_delivers_its_related_cards, test_ebays_fewer_words_cards_are_never_billed |
| 18% of runs timed out (one scraper's 30-day stats) | The run stops before its timeout, delivers what it has and closes every browser | test_near_the_run_timeout_no_new_target_or_page_starts_and_the_run_ends_green, test_no_browser_outlives_the_run |
| Billed past the limit | Charged equals delivered, and the spending limit is exact | test_charged_never_exceeds_delivered_in_any_mode, test_sdk4_spend_cap_is_exact_with_details |
What users of other eBay scrapers praise, and how this actor measures up
From the 4 and 5-star reviews of the other eBay scrapers on Apify (82 reviews, read 2026-11-18 and re-read 2026-11-19). Each line on the right is backed by a named offline test or a dated run.
| What reviewers praise | This actor | Proof |
|---|---|---|
| "Simple setup and beginner-friendly", "works with minimal input" | One keyword is a whole input ({"queries": ["mechanical keyboard"]}), the form starts filled in, an empty input ends free with an example of the smallest one, and the Quick start above has the Console, API and agent routes | test_a_keyword_delivers_rows_and_charges_exactly_them, test_an_empty_input_shows_the_smallest_input_that_works |
| "Reliable for daily tasks", "I can reliably set it and forget it" | Sessions that passed eBay's bot wall are kept for your next runs; a refused page is an error row, never an empty success, and the run card says what to do | Between 2026-11-18 and the morning of 11-19, 159 of the 160 runs we started and let finish succeeded; the one failure (kjj42VqM8oBtlltVm) was refused at every browser visit for 7 minutes. test_a_refused_run_says_what_to_do |
| "Data extraction is quite thorough", "Rich data fields" | 43 columns on every search row, units sold and watchers included; item pages add specifics, gallery, variations, identifiers, return policy and parts compatibility | test_every_item_row_is_complete, test_item_urls_carry_the_details_without_the_add_on_charge |
| "Grabbed all the data I needed in just a few minutes" | 60 listings 4 to 6 s after the start once a session is kept; 32 s to 2.6 minutes to the first page on a first run, with the progress on the run card meanwhile | runs 9AlfLmNXn9FrPRk81, RORZbbny5sKaPesYs, 2jnbl7CpEohVaixeq, GKsEigVOoOk1cw4m7, MjRnCVUCsld10DDfb, koWunted0cGQ6CzNq; test_the_run_card_shows_progress_while_the_run_works |
| "The process is transparent, and it's easy to track the progress" | Every few seconds the run card says the listings so far, the jobs done, the pages read and how eBay's bot check is going, and why a finished run is still waiting; at the end, what was left out free and why | run koWunted0cGQ6CzNq (seven progress lines through six refused browser visits); test_the_run_card_shows_progress_while_the_run_works, test_the_run_card_says_when_a_done_run_keeps_its_session |
| "Turns raw data into actionable pricing advice in seconds" | A free market summary on every search (price quartiles, by condition and format, units sold, top sellers, cheapest), with quick_take, the market in a few sentences | test_the_market_summary_says_the_market_in_plain_sentences; runs zNOfisgvq1Oz1Auac, ZyNkPks78nd2FcDsT |
| "Handles complex pagination perfectly" | Past eBay's ~2,000 per search in price bands, page by page, with excludeDatasetIds to carry on | run 0KreSWCwonduQxqIV (2,100 of 2,100); test_a_deep_walk_returns_every_listing_once_past_ebays_stop |
| "Everything works and bills exactly as documented", "good value for money" | Charged equals delivered; status rows, filtered listings, duplicates and price-band probes are free; the spending limit is exact | test_charged_never_exceeds_delivered_in_any_mode, test_sdk4_spend_cap_is_exact_with_details; run yTOYEJE6BxBSmpmIh |
| "The API is very convenient to call", exports to CSV, Excel and Sheets | One run-sync-get-dataset-items call returns CSV; statusRows: "none" keeps the dataset to listings (43 flat columns, where a full export of a run with its status row had 127: RORZbbny5sKaPesYs); every view names its columns in words; recipes for MCP agents, Google Sheets, Make, n8n, Zapier, Slack and webhooks | runs KcVLqzGRWYQ0g5EnK (none: 20 listings, no status row) and NGaetw2P8PhdwPSpf (problems: 323 listings, no status row since nothing went wrong); test_status_rows_all_problems_or_none, test_every_view_has_readable_column_names_and_the_overview_links_each_listing |
| "Gives sold data" | ✗ eBay shows sold search only to signed-in accounts; such a request ends free with the reason, and the auctions you monitor come back with their final price (add-on, once its price is set) | test_a_sold_request_ends_free_with_the_reason_and_fetches_nothing, test_sold_auctions_are_read_once_after_they_end_and_billed_as_sold_items |
Questions users ask about other scrapers
The most common complaints in the issues and reviews of the other eBay scrapers on Apify (82 reviews and 56 issue threads, read 2026-11-18), and what this actor does. Each answer names the run or the offline test that shows it.
Will a blocked run come back as "Succeeded, 0 items"? No. An eBay refusal is an error status row and a run where everything was refused fails, so a schedule or an agent can tell a block from an empty market (test_a_refusal_is_reported_as_an_error_not_as_no_results, test_every_job_failing_fails_the_run). No API token is involved that could expire. 14 of 14 reproduction runs on 2026-11-18 succeeded, four store monitors started at once among them (XnZNb3ysCwkggsBkH, Cbn6iuvQCmK5ErZky, adIL55OvOpJNyRHFy; the fourth met eBay's false empty page, see the next answer).
My run said 0 results, but the seller has listings. Why? eBay sometimes answers one IP with an empty page: a seller with 96 listings came back empty once in 21 seller searches on 2026-11-18 (cptr3Uw0dzibxXBtw), and the same URL gave 96 minutes later. Since 1.10 an empty first page of a seller, a Store or a monitored search is read again on another IP before it counts as nothing found, and the status row's note says what the second read found (test_an_empty_seller_page_is_read_again_before_it_counts_as_nothing).
Are UK, Australian and German prices the ones a local buyer pays? Yes, since 1.10. eBay prices each page for the visitor's own country, and a US exit saw ebay.co.uk private sellers £0.47-0.98 cheaper than a UK buyer does (wkKkn1jALejlqHFpA against 7ASD7hmpUWvAlEZr1) and lost Australian postage (Lt7ZWYTT7Dnx0YCaN against PZ3vJKHjVHhFGYFnA). Every exit is now in the marketplace's country and every page's ship-to code is checked: the same inputs gave 16 of 16 UK prices equal to a UK buyer's (48bVEoqE77K9UwaoQ) and 17 of 17 Australian rows with their postage (ycrhSP9PYVY7BGAfN; test_a_page_priced_for_a_buyer_abroad_moves_the_run_to_residential_in_the_country).
Do I pay for near-matches and "results matching fewer words"? Not if you say so: titleMustMatch leaves out titles without every keyword word, free (memo23's own input billed 20 near-matches on 1.9.2, run njtQYO6xWW9fiAwvn, and 0 with it, run w0YUVoeJJAPCoX2ah). eBay's "fewer words" padding is never delivered, with or without it (test_only_ebays_fewer_words_notice_cuts_a_page).
Can I monitor many stores every day and pay only for new listings? Yes. Nine sellers and an item URL ran in one run in 61 s (90ZVCUPuwTjkoVjz6); with onlyNew each later run bills only listings it has not delivered before, and blackfalcondata's incrementalMode, stateKey and emitUnchanged are read as ours: its own store-monitor input, run twice, billed 10 new listings each time and skipped the 9 the first run had delivered (jOUBlB7uS0iSxxwIJ, 7z0PGYlosQsXxtSz1).
Why do I get fewer listings than eBay says it has? eBay pages through about 2,000 listings of one search. Above 2,040 the search is cut into price bands that eBay shows in full: 2,100 of 2,100 unique listings in run 0KreSWCwonduQxqIV (test_a_deep_walk_returns_every_listing_once_past_ebays_stop). eBay also "streamlines" some searches, leaving matching listings out of its default view (249 of 367 in one band on 2026-11-18): a walk reads the full view, and a plain search that eBay streamlined says so on its status row, with how to get every listing (test_a_streamlined_search_says_so_on_its_status_row_and_the_run_card).
My big job stopped at its time limit. Do I lose it, or pay twice to finish it? Neither. A 2,500-listing walk stopped at its 300 s limit kept its 1,380 listings and printed the value to carry on (l6SLUJnmhzRUhZj2B); the second run delivered the other 1,122 and billed none of the first 1,380 again (Rq7Mo0kmUUpBzAi2e). A walk delivers and bills page by page, so what was read is kept (test_a_band_walk_delivers_page_by_page_and_keeps_what_it_read_when_a_page_is_refused); the run card gives the excludeDatasetIds value that carries on while skipping, free, what was delivered (test_a_band_walk_cut_by_the_run_timeout_says_how_to_carry_on, test_exclude_dataset_ids_carry_on_without_paying_twice); a run the platform restarts never bills a listing twice (test_a_restarted_run_never_delivers_or_bills_twice).
My filters were "accepted without error and silently ignored". Here too? No. Other scrapers' names for the marketplace, the counts, the filters and the monitor settings are read as ours and listed in SUMMARY.aliased_fields; caffein.dev's keyword, marketplace: "EBAY_GB" and count gave 20 rows in pounds from ebay.co.uk (wkKkn1jALejlqHFpA), and their limits and money-saving filters are never ignored: automation-lab's, apt_marble's and memo23's fields in one input reached eBay as LH_FS=1&LH_PrefLoc=1&Brand=Logitech, and 17 cards under memo23's watcher floor were left out free (12ihMxBqu60pUfQ2f; test_other_scrapers_limits_and_filters_are_never_ignored). A name nobody uses is listed in ignored_fields and on the run card.
Does it return sold prices? Sold and completed search needs a signed-in eBay account, so a sold request ends at once, free, with the reason (test_a_sold_request_ends_free_with_the_reason_and_fetches_nothing). What it can follow signed out are the auctions a monitored search delivered: each is read again once it has ended (the sold-auctions add-on, active once its price is set; 6 sold of 20 read in test run NYzPp9BeFcYCxtbN8).
Does it respect my spending limit? Yes, exactly: charged always equals delivered and the cap is never passed (test_charged_never_exceeds_delivered_in_any_mode, test_sdk4_spend_cap_is_exact_with_details, test_sdk4_spend_cap_is_exact_with_a_title_rule).
Does a run hang until its timeout? No. It stops starting new pages before the timeout, delivers what it has, closes every browser and ends green (test_near_the_run_timeout_no_new_target_or_page_starts_and_the_run_ends_green, test_no_browser_outlives_the_run).
Seller feedback: what buyers say about a seller (add-on)
feedbackSellers takes eBay seller usernames or feedback-profile URLs (https://www.ebay.com/fdbk/feedback_profile/<seller>; such a URL in startUrls works too). Each seller's public feedback profile is read signed out, newest first (feedbackSort: "most_relevant" for eBay's own order), 200 entries a page, and every entry becomes a feedback row:
feedback_rating(positive,neutral,negative) andcomment;left_by:buyer(feedback the seller received from a buyer),seller(received as a buyer) orebay(eBay's automated feedback), withfrom_user(masked, e.g.n***s) andfrom_user_score;when(eBay's own bucket: "Past month", "Past 6 months" …),verified_purchase,image_countandimages;- the item it was left for:
item_id,item_title,item_price,item_currency,item_url.
maxFeedbackPerSeller (default 100, up to 10,000), feedbackRating and feedbackReceivedAs (seller / buyer) choose what you get; entries that do not match are left out free and counted on the seller's status row, whose note carries the seller's feedback score and the last 12 months' positive and negative counts. eBay's profile page does not apply rating filters from its URL, so they are checked on each entry. memo23's feedbackProfiles, feedbackRating, feedbackReceivedAs (both = all), feedbackSort, feedbackImagesOnly and feedbackExcludeAutomated work as they do there, and its maxItems is the count per seller (test_memo23s_feedback_input_runs_unchanged).
This is an add-on billed per feedback row once its price is set; until then a run with it reads no feedback, charges nothing for it and says so on the run card (test_seller_feedback_is_gated_until_its_price_is_set). Billing follows the same rules as listings: one event per delivered row, never past your spending limit, and never twice after a restart (test_seller_feedback_rows_are_billed_one_event_each_up_to_the_cap, test_sdk4_spend_cap_is_exact_with_seller_feedback, test_a_restarted_run_never_bills_feedback_twice).
Seller and store listings
sellers takes eBay usernames or links to a seller's profile or eBay Store. On its own it lists everything each seller has active (one job per seller, 60 listings a page, pages deep). With queries, every keyword is searched within those sellers only (eBay's own "only these sellers" search, one search per keyword whatever the number of sellers). A Store's name is often not its seller's username (australiancomputertraders belongs to bneacttrader), so a Store link costs one extra page read, which is free. Profile (/usr/<name>), Store (/str/<name>) and seller-items (/sch/<name>/m.html) links also work in startUrls. A seller with nothing listed on that marketplace is a free no_results row.
{"queries": ["iphone 13"],"sellers": ["verizon", "vipoutlet"],"domain": "com"}
Monitoring: only new listings and price changes
For a scheduled run, set onlyNew: true. The run then delivers, and bills, only listings that this exact search (same keyword or URL, marketplace and filters) has not delivered before. Listings seen again are skipped free and counted as already_seen on the status row and in SUMMARY. The first run delivers everything it reads, so start it on the day you want the baseline. Add includePriceChanges: true to also get listings whose price moved since they were last delivered; each row then carries change (new or price_change) and previous_price.
The memory is the key-value store ebay-search-seen in your own Apify account: the newest 20,000 listings per search, with their last price, and for auctions their end time. The actor creates that store on its first monitoring run; delete it to start over. pages decides how deep each run looks, and maxResultsPerQuery and maxItems count the new rows a run delivers, not the rows it reads. Pair it with sortBy: "newly_listed" for new-listing alerts.
includeSoldAuctions: true (it turns onlyNew on by itself) also tells you what the auctions this search delivered sold for. Once an auction's end time has passed, its own page is read once more: eBay shows an ended listing's result to anyone, signed out (10 of 10 ended listings read on 2026-11-18). A sold one is delivered as a row with change = sold: price is the final price, bids the bids it drew, ended_at when it ended ("Nov 17, 2026 23:11:13 PST"), and previous_price its price when it was delivered (left empty when the ended page shows another currency: a UK listing's page shows pounds where a US search card showed dollars). An unsold one is noted in SUMMARY and not delivered, and each is read at most once (up to 100 a run). Buy It Now listings are not followed: one that leaves a search page has usually just moved down it. This is an add-on billed per sold auction; until its price is set, a run with it reads nothing, charges nothing for it and says the add-on is not active yet.
blackfalcondata's monitor input runs here unchanged: incrementalMode: true turns on onlyNew and price changes ("new or changed listings"), stateKey names the memory, so a monitor keeps it when you change a filter and two monitors of the same search can keep separate ones, and emitUnchanged: true also delivers the listings seen before, as change = unchanged (billed like any row). An empty first page of a monitored search that has delivered before is read again on another IP before the run reports that nothing is new.
How much does it cost?
Pay per event: $0.004 per listing row ($0.0035 on Scale, $0.003 on Business), plus $0.0025 per listing enriched by the item-details add-on ($0.0022 / $0.002) when you turn it on (from 2026-12-02). There is no per-run start fee and no compute charge on top.
- A search that returns nothing is free.
- Listings where eBay withholds the price ("See price") are dropped free whenever you have set a price bound.
- Rows dropped by your price bounds are free, and so are titles without every keyword word (
titleMustMatch), eBay's "results matching fewer words" padding, and cards that failfreeShipping,itemLocation,excludeWordsor another scraper's filter you passed (minWatchers,minQuantitySold,sellerCountry…). The status row'sleft_outcounts them by reason. - Listings already in the datasets you list in
excludeDatasetIdsare skipped, free, and a run the platform restarts never bills a listing twice. - A listing found by two of your keywords is delivered and billed once per run.
statusrows are free.- Price-band probe pages (a band's first page read only to cut it in two) deliver nothing and cost nothing.
- The sold-auctions add-on (
includeSoldAuctions) is billed per sold auction once its price is set; until then it reads nothing and costs nothing. - If your spending limit is reached mid-run, the run stops delivering rather than delivering rows it could not charge for — you never receive unpaid data and never pay for data you did not receive.
One page of 60 listings is $0.24. Five keywords × two pages × 60 delivered listings × $0.004 is at most $2.40 per run, or $72 for 30 daily runs at the current rate.
Every input, explained
| Input | Default | What it does |
|---|---|---|
queries | [] | Keywords to search, one job each |
sellers | [] | Seller usernames or Store/profile URLs: alone, each seller's listings; with queries, keywords searched within these sellers |
startUrls | [] | Active eBay search URLs (their filters are kept) and item URLs (one item-page row each) |
domain | com | Marketplace for keyword and seller searches (15 hosts, co.uk, ca, ie, at, ch included); ebay.com, www.ebay.de, US, UK work too |
buyingFormat | all | all, buy_it_now, auction or best_offer (listings that accept offers) |
conditions | [] | Multi-select of the nine eBay condition grades; refurbished = all five refurbished grades |
minPrice | 0 | Lower bound (decimals allowed), sent to eBay and re-checked before billing |
maxPrice | 0 | Upper bound (decimals allowed); 0 means none |
category | "" | eBay category id, or a category page / category search URL (its id is read, and its marketplace when domain is left at its default) |
sortBy | best_match | best_match, ending_soonest, newly_listed, price_low, price_high |
excludeWords | [] | Words or phrases to leave out of the titles: words go to eBay as -word, phrases are dropped before billing |
titleMustMatch | false | Only titles with every word of the keyword; near-matches are left out free and maxResultsPerQuery counts matches |
freeShipping | false | Only listings with free shipping (eBay's filter, re-checked on each card) |
returnsAccepted | false | Only listings whose seller accepts returns (eBay's filter) |
itemLocation | default | domestic = located in the marketplace's own country (eBay's filter, re-checked on each card), worldwide |
pages | 1 | Result pages per job, 60 listings each, up to 34 |
maxResultsPerQuery | 0 | Cap per job; 0 means everything the pages returned; above 2,040 (up to 100,000) the search is cut into price bands |
includeDetails | false | Add-on: each delivered search listing enriched with its item page (description, gallery, variations, identifiers, seller); $0.0025 extra per enriched listing, from 2026-12-02 |
onlyNew | false | Scheduled runs: only listings this search has not delivered before |
includePriceChanges | false | With onlyNew: also listings whose price changed, with change and previous_price |
includeSoldAuctions | false | With onlyNew (turns it on): auctions this search delivered come back once they have sold, with change = sold, the final price and ended_at (add-on, active once its price is set) |
searchInDescription | false | eBay's "Search including description": keywords also match listing descriptions |
marketSummary | true | Free market summary on each search's status row (market) and in SUMMARY.markets |
includeDescriptionHtml | false | Item-page rows also carry description_html, the seller's description markup (free) |
allCompatibleVehicles | false | Item-page rows of parts list every compatible vehicle, not just the first 20 (free) |
feedbackSellers | [] | Add-on: sellers whose feedback profile to read, one feedback row per entry (active once its price is set) |
maxFeedbackPerSeller | 100 | Feedback rows per seller (0 = up to 10,000) |
feedbackRating | all | positive, neutral or negative only (checked on each entry, free) |
feedbackReceivedAs | all | seller (from buyers, eBay's automated feedback included) or buyer |
feedbackSort | most_recent | most_recent or most_relevant |
statusRows | all | all = one free status row per keyword, URL or seller; problems = only those that failed, found nothing, stopped at a limit or carry a note; none = listings only, for clean CSV and Excel exports (the STATUS record and SUMMARY keep every one) |
excludeDatasetIds | – | Datasets of earlier runs: listings already in them are skipped, free (carry on a big search without paying twice) |
sessions | 4 | Parallel proxy sessions (egress IPs) |
perIp | 0.5 | Requests per second per IP |
reuseSessions | true | Keep the sessions that passed eBay's wall for your next runs (store ebay-search-sessions in your account); false = every run starts fresh |
proxyConfiguration | Apify Proxy | Required; the run escalates to residential on its own |
maxItems | – | Optional total cap on listing rows for the whole run (the name AI agents send); it also lifts pages when that was left at 1 |
Inputs from other scrapers work too
An input written for another eBay scraper runs here without edits. Its field names are read as ours only where our own field is not set, and every name that was read this way is listed in the run's SUMMARY record as aliased_fields. Anything still unknown is ignored rather than failing the run, and is listed as ignored_fields and at the end of the status message.
| their field | read as |
|---|---|
keywords, keyword, search, searchQuery, searchQueries, query, searchTerms | queries (merged with ours) |
listingUrls, Urls, urls, url, startUrl, itemUrls, productUrls; itemIds | startUrls (merged; item ids become item URLs) |
ebaySite, marketplace, site, country ("ebay.com", "www.ebay.de", "US", "UK", "EBAY_DE" …) | domain |
count, maxResults, maxProductsPerSearch, maxItemsPerQuery, max_results | maxResultsPerQuery, and the pages needed to reach it (price bands above 2,040) |
limit | maxItems |
splitByPrice: true | every price band (maxResultsPerQuery 100,000) |
maxPages (0 = every page), maxSearchPages | pages |
condition, itemCondition, conditionId ("new", "used", "refurbished", "For parts or not working", 3000 …) | conditions |
listingType, buyItNowOnly ("bin", "buyItNow", "LH_Auction", "acceptsOffers" …) | buyingFormat |
sort, sortOrder ("price_asc", "newest", "pricePlusShippingLowest", eBay's _sop numbers …) | sortBy |
categoryId, subcategoryId (caffein.dev: the subcategory wins), category_id ("0" = all), memo23's motorsCategory ("none" = no preset) | category |
includeSummary (memo23) | marketSummary |
feedbackProfiles, feedbackImagesOnly, feedbackExcludeAutomated (memo23; its feedbackRating, feedbackReceivedAs and feedbackSort are our names too) | feedbackSellers and the feedback filters |
searchTitleDescription (delicious_zebu) | searchInDescription |
splitByPrice: false | read: the default, one search |
min_price, max_price, priceMin, priceMax | minPrice, maxPrice |
max_items | maxItems |
seller, sellers (a list, or "a, b" in one string), storeUrls | sellers |
monitoringMode, onlyNewItems, incrementalMode (blackfalcondata: also price changes) | onlyNew |
stateKey, emitUnchanged (blackfalcondata) | the monitor's memory name; also deliver unchanged listings |
titleMustMatchQuery; titleMustInclude (your own words) | titleMustMatch |
excludeFewerWordMatches, exactMatch | always on: eBay's "fewer words" padding is never delivered |
freeShippingOnly | freeShipping |
homeCountryOnly; caffein.dev's, memo23's and khadinakbar's itemLocation | itemLocation |
minWatchers, minQuantitySold, sellerCountry, minYear, maxYear | free checks on each card before billing |
sellerType, aspectFilter ({"Brand": "Nike"}), keywordMatch, locatedIn | eBay's own filters (LH_SellerType, item specifics, _in_kw, LH_LocatedIn) |
query as a JSON array ('["switch oled", "ps5 slim"]') | several keywords |
enablePriceMonitoring | includePriceChanges |
detailedItems, includeDescription | includeDetails |
excludeKeywords (a list, or one string of words), negativeKeywords, excludedKeywords | excludeWords (merged with ours) |
proxy, proxyConfig | proxyConfiguration |
One name means different things in different scrapers: in memo23's and epicscrapers' inputs maxItems is a cap per search, in ours it is a cap for the whole run. We read it per search only when the input is clearly shaped for those scrapers, which means it carries one of their own names (detailedItems, monitoringMode, maxConcurrency, site and similar) and does not set our maxResultsPerQuery.
A request for sold or completed listings (mode: "sold" or "both", daysToScrape, soldWithinDays, includeCompletedListings: true, sortOrder: "endedRecently", or a sold search URL) is not converted into an active search: it ends at once with the reason and costs nothing. mode: "active" is fine.
How do I call it from Python, JavaScript or curl?
Python:
from apify_client import ApifyClientclient = ApifyClient("<YOUR_APIFY_TOKEN>")run = client.actor("kestrel/ebay-search-scraper").call(run_input={"queries": ["mechanical keyboard"],"pages": 2,"minPrice": 30,"maxPrice": 200,})for row in client.dataset(run["defaultDatasetId"]).iterate_items():if row["type"] == "item":print(row["price"], row["currency"], row["units_sold"], row["title"])
JavaScript (Node):
import { ApifyClient } from 'apify-client';const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });const run = await client.actor('kestrel/ebay-search-scraper').call({queries: ['keycap set'],pages: 1,sortBy: 'price_low',});const { items } = await client.dataset(run.defaultDatasetId).listItems();console.log(items.filter((r) => r.type === 'item').length, 'listings');
curl:
curl -X POST "https://api.apify.com/v2/acts/kestrel~ebay-search-scraper/run-sync-get-dataset-items?token=$APIFY_TOKEN" \-H 'content-type: application/json' \-d '{"queries":["vintage pocket watch"],"buyingFormat":"auction","pages":1}'
Integrations: AI agents (MCP), Google Sheets, Make, n8n, Zapier, Slack and webhooks
AI agents through MCP. Apify's MCP server offers this actor as a tool to Claude, ChatGPT, Cursor, VS Code and any MCP client.
- Claude Code:
claude mcp add --transport http apify "https://mcp.apify.com?tools=kestrel/ebay-search-scraper" - Claude Desktop, Cursor or VS Code, in the MCP config file (your browser opens once to sign in to Apify):
{"mcpServers": {"apify": { "url": "https://mcp.apify.com?tools=kestrel/ebay-search-scraper" }}}
Then ask in plain words: "Find the 20 cheapest new Sony WH-1000XM5 on eBay UK with free shipping", "Which sellers dominate used iPhone 13 listings on eBay Germany, and at what prices?", "What does the eBay market for Lego 42143 look like?". The agent gets typed rows; each search's status row says what it found, left out and charged, market.quick_take sums the market up in a few sentences, and a refused run fails with the reason instead of returning an empty success.
Google Sheets. In a cell: =IMPORTDATA("https://api.apify.com/v2/acts/kestrel~ebay-search-scraper/runs/last/dataset/items?format=csv&view=overview&status=SUCCEEDED&token=<YOUR_APIFY_TOKEN>"). That is the Overview table (title, price, shipping, condition, units sold, watchers, seller, link) of your last successful run, refreshed by the sheet. Anyone who can open the sheet can read the token, so for a shared sheet use Apify's Google Sheets integration instead (a saved task's Integrations tab: append each run's rows to a sheet).
Make. Apify app: "Run an Actor" (actor kestrel/ebay-search-scraper, input JSON as in the examples, run synchronously), then "Get Dataset Items" with the run's defaultDatasetId, then for example Google Sheets "Add a Row" or an email.
n8n. The Apify node: run the actor kestrel/ebay-search-scraper with the input JSON, then read the dataset items of that run; the rows come back as n8n items, one per listing (add "statusRows": "none" to skip status rows).
Zapier. Apify app: "Run Actor" as an action, or "Finished Actor Run" as the trigger of a saved task you schedule in Apify, then "Fetch Dataset Items".
Slack, email and your own endpoint. On a saved task's Integrations tab, Slack and email post when a run succeeds or fails, and an HTTP webhook on "Run succeeded" POSTs the run to your URL; its resource.defaultDatasetId is the dataset to read. Schedule the task with onlyNew: true and each message is about the new listings only.
Your own code. See Python, JavaScript and curl above; format=csv, format=xlsx and view=overview (or demand, sellers, price_research, monitoring) work on every dataset-items endpoint.
eBay API alternative: why not use the official API?
The eBay Browse API is excellent if you are already an eBay developer: you register an application, complete the onboarding, manage OAuth tokens and live inside its rate limits and its field set. For a one-off market study, a price watch, a sourcing spreadsheet or an agent tool, that is a lot of ceremony. This actor is the ebay api alternative for those cases — paste keywords, get rows, pay for what you got. It also returns things the catalogue-shaped API does not surface the same way: the struck-through list price a seller is advertising against, the watcher count, the units already sold on that specific listing, and the exact position the listing held in eBay's own results.
If you need order management, listing creation, or guaranteed contractual access, use the official API. If you need to read the public market, use this.
Does this scraper collect eBay sold items?
No, and here is why. eBay's completed and sold search — the LH_Sold=1&LH_Complete=1 view — is behind a sign-in gate. A logged-out request to it is redirected to signin.ebay.<tld> with sgfl=srch, on every marketplace. We verified this on 2026-08-29 across seven marketplaces, datacentre and residential proxies in seven countries, eight different URL routes including eBay's own "Sold items" rail link and the RSS feed, and three browser fingerprints: not one of roughly 115 attempts returned sold results. We re-checked on 2026-11-17 with a real browser that eBay had just let through, on ebay.com, .co.uk, .ie, .ch and .at, from a home connection and from residential proxies: LH_Sold=1&LH_Complete=1, and LH_Complete=1 alone, went to the sign-in page every time. Only a signed-in eBay account gets through, which we will not do on your behalf or ours.
So this actor searches active listings only. What it can do signed out is follow the auctions it found: an ended listing's own page still shows whether it sold, for how much and when, to anyone. With onlyNew and includeSoldAuctions (see Monitoring), every auction a scheduled search delivered comes back once with its final price after it ends. That covers auctions you are watching, not eBay's whole sold history. A request for sold listings (a sold search URL, mode: "sold", a sold-days window such as daysToScrape, or the "ended recently" sort) ends at once with an explanation and costs nothing; it is never answered with active listings you did not ask for. If you came here for ebay sold items to build price comps, the honest substitute is the units_sold column: it is the number of units a live listing has actually sold at its current price, published by eBay on the public results page, and for pricing decisions it answers most of the same question — has anyone paid this? — using data that is genuinely available.
FAQ
How fast is it?
A first run in your account waits while a real browser gets past eBay's bot check, about 20 s per browser visit: on 2026-11-19 first runs needed one to seven visits, and their first page came 32 s to 2.6 minutes after the start (runs 2jnbl7CpEohVaixeq, GKsEigVOoOk1cw4m7, MjRnCVUCsld10DDfb, koWunted0cGQ6CzNq). The sessions that passed are kept for your next runs, which need no browser: 60 listings 4 to 6 s after the start (9AlfLmNXn9FrPRk81, RORZbbny5sKaPesYs). While it waits, the run card says so ("eBay's bot check: real-browser visit 3 on a fresh IP, 2 refused so far"; run koWunted0cGQ6CzNq showed seven such lines). After its last listing, a run whose browser is still loading waits up to 25 s for it to hand over its cookies, so that your next runs need no browser, and its run card says that too.
How do I get a clean CSV or Excel file?
Set statusRows: "none" and export: the dataset then holds listings only, 43 flat columns. With the default, each search also adds one free status row, and a full CSV export of a run with one search had 127 columns, 75 of them from its market summary (RORZbbny5sKaPesYs). Or export a view: the Overview table (title, price, shipping, condition, units sold, watchers, seller, link) is ?format=csv&view=overview on the dataset's items URL, and fields=title,price,currency,url picks columns yourself. Excel is format=xlsx.
My run failed. What should I do?
Read the run card: a run fails only when eBay refused every page it asked for (its bot check), and then nothing was charged. Run it again in a few minutes: the next run starts on fresh IPs and keeps the ones that pass for your later runs. If it keeps happening, set proxyConfiguration to Apify's RESIDENTIAL group. A search that finds nothing, a bad input and a sold request do not fail: they end green, free, with the reason.
How many listings are on each eBay page?
Sixty. The actor requests _ipg=60, which is eBay's ceiling. pages: 3 therefore reads up to 180 listings per job. eBay stops paginating any single search at roughly 2000 items however many it reports in results_total, which is why pages is capped at 34; for more, see the price bands above.
Why did my search return no rows?
Either eBay genuinely found nothing, or the run was refused. The two are never confused: a genuine miss produces a status row with status: "no_results" and costs nothing, with a note on what to try (the spelling, a looser filter, another marketplace), and a refusal produces status: "error" with the reason and what to do. eBay also sometimes answers one exit with an empty page for a search that has listings (once in 21 seller searches on 2026-11-18): an empty first page of a seller, a Store or a monitored search that has delivered before is therefore read again on another IP, and the status row's note says what the second read found. The run card counts the searches that found nothing. This matters more than it sounds — eBay's zero-result page still renders about twenty "results matching fewer words" cards from other searches, and a scraper that delivers those bills you for listings you never asked for. This one drops them.
Do I need a proxy?
Yes, and Apify Proxy is the default. eBay is behind Akamai Bot Manager, which since 2026-11-13 refuses the first request of any plain HTTP client. Each proxy session therefore opens with one real-browser (Camoufox) page on its own IP; the pages after it reuse that browser's cookies on the same IP, which is what a visitor's browser does. A refused session is closed and replaced by a fresh IP, and the run escalates from datacentre to residential proxies by itself once eBay has pushed back twice. In cloud runs on 2026-11-16 and 11-17 a browser page passed on about 1 in 5 datacentre IPs and 2 in 5 residential IPs; a refused attempt costs a few seconds and nothing else. The first page comes back as soon as eBay has sent it, so a one-page run does not wait for the browser to finish.
Since 1.8 the sessions that passed are kept in the key-value store ebay-search-sessions of your own Apify account: the proxy session id, the IP it leaves from, eBay's anonymous cookies and the browser's User-Agent. Your next run reopens them (Apify keeps a datacentre session's IP for 26 hours after its last use and a residential one's for about 30 minutes; the actor keeps them 24 hours and 20 minutes), asks an IP echo which IP the session now leaves from, and replays the cookies only when it is the same IP. A reopened session needs no browser; if eBay has expired its cookies, the session gets one browser page on the same IP before a fresh IP is tried. Two of your runs at the same time never share a session. At the end of a run whose browser is still loading, the actor waits up to 25 s for it to hand over its cookies (on Apify a residential page took 20-21 s, 2026-11-18); past that it keeps the session's IP without cookies, and the next run opens one browser page on that proven IP instead of trying random ones. Set reuseSessions: false to start every run fresh.
Are sponsored listings flagged?
There is deliberately no sponsored column. eBay's current results page prints the (deliberately obfuscated) "Sponsored" disclosure on every card — 60 of 60 on four different searches across two marketplaces and both buying formats, all from the same organic results module. A column that is always true is worse than no column, so we left it out and documented the finding instead.
Can I track prices over time?
Yes — schedule the run and keep the datasets. Each row carries fetched_at, item_id and price, which is everything a time series needs. To pay only for what changed, set onlyNew and includePriceChanges (see Monitoring above). A daily run of the same keywords, appended to one table and grouped by item_id, gives you an ebay price tracker with the listing's units_sold and watchers moving alongside the price, which is far more informative than the price alone.
What about eBay Kleinanzeigen or eBay Motors?
Kleinanzeigen is a separate site with its own markup and is not covered. Motors search listings use the same search-row format, and vehicle attributes (year, make, model, mileage) are in the item_specifics of an item-URL row.
Is scraping eBay legal?
This actor reads only publicly available pages — the same search results and item pages any visitor sees without logging in — and collects no personal data beyond the seller usernames and store names eBay itself publishes next to every listing. Public-data scraping of this kind is broadly lawful in the US and the EU, and courts have repeatedly declined to treat access to a public web page as unauthorised access. That is not the whole story, though: eBay's own Terms of Service restrict automated access, so scraping can be a contractual matter between you and eBay even where it is not a legal one. You are responsible for your own use — respect robots directives and rate limits, do not republish listing content wholesale, keep collection proportionate to your purpose, and take your own legal advice for anything commercial. We do not scrape anything behind a login, and we will not: that is exactly why the sold-listings view above is absent rather than faked.
How do I find an eBay category id?
Open any eBay category in a browser and read the _sacat number out of the URL, or click into a category from a search and copy it from the address bar. Pass it as category and every keyword in the run is searched inside that category only.
Changelog and support
See CHANGELOG.md for version history. If a field stops populating, open an issue on the actor's Issues tab with the input you used — the parsers are covered by an offline test suite built from captured pages, and a fixture from a real failing run is the fastest possible fix.
Related scrapers
eBay is one marketplace. These read the others, with the same row-per-item shape and the same pay-per-delivered-row billing:
- Amazon Product Scraper — one ASIN in, the whole product page out: buybox and list price, stock flag, gallery, bullet points and the details table, across 20 marketplaces — the Amazon column of a price-spread table.
- Amazon Search Scraper — Amazon keyword results with page position and a sponsored flag, when you want to know what ranks for a phrase there the way this actor tells you for eBay.
- Amazon Best Sellers Scraper — the Top 100 of any Amazon category with every rank filled, plus New Releases, for demand ranking rather than price.
- Amazon Reviews Scraper — customer reviews and ratings by ASIN with filters that run before billing, for the feedback behind a listing you priced here.
- AliExpress Reviews Scraper — buyer feedback with photos, the variant chosen and the buyer's country, for checking a supplier's product before you list it on eBay.
All of them bill per delivered row, never charge for rows a filter or a spending limit removed, and write an Apify dataset you can export to CSV, Excel or JSON.