Whatnot Live Shopping Scraper

Whatnot is the largest live-shopping marketplace for collectibles: trading cards, sports cards, sneakers, comics, coins, toys and more. Sellers run scheduled live shows and auction or sell items inside them.

This actor turns that into structured data. It returns live shows (playing now and scheduled for later), the items and lots sellers have listed, and seller profiles, with the site's own category, format, seller-rating and country filters, plus resume and incremental change tracking for scheduled runs.

What is not available, stated up front

Realised (sold) prices for finished shows are not published by the site, so this actor does not return them. This is measured, not assumed:

• The site's own filter for a show's time-of-show offers exactly two values, "Playing now" and "Scheduled". There is no ended or completed value, so finished shows are not in the public index at all.
• A show carries no items connection, so there is no path from a show to what it sold.
• A seller's own show list returns only future shows.
• The catalogue product type declares a last-sale price field, and it came back empty on every one of 24 distinct products sampled.

If you need realised auction prices, this actor cannot give them to you and neither can any other public read of this site. What it does give you is the pre-sale picture: which shows are running and scheduled, who is running them, how big their audience is, and exactly which lots are queued into which upcoming show at what starting price. That is the data you use to decide what to watch, what to bid on, and which sellers to track.

What you get

Live shows - title, description, status (playing now or scheduled), start time, current viewer count, watchlist size, categories, tags, free-shipping policy, the maximum shipping the buyer pays, thumbnail, trailer, and the full seller block.

Items and lots - title, subtitle, description, price and currency, buying format (Buy It Now or live auction), current bid and bid count, quantity, listing status, photos, condition-bearing category, catalogue product and its price range, break details (format, total spots, spots filled, spots sold), variants, listing and update timestamps, and the start time of the show the lot is queued into.

Sellers - username, display name, bio, rating and review count, followers, following, items sold, average shipping days, verified flag, Premier Shop flag, live flag, store banner, and optionally their most recent reviews with rating, text, seller reply and date.

Every show and every item also carries its seller's username, rating, review count, follower count, sold count, verified flag and Premier Shop flag, so you can rank by seller quality without turning anything extra on.

Modes

Search - one or more keywords. Pick what each search returns with Result type: live shows, items and lots, sellers, or the platform's own mixed feed. Each keyword gets its own share of the result cap, so one popular keyword cannot swallow the run.

URL - paste links and each is routed automatically:

Link	What you get
whatnot.com/live/<id>	that one show
whatnot.com/listing/<id>	that one item
whatnot.com/user/<username>	the seller, plus their scheduled shows and their shop
whatnot.com/tag/<category>	shows in that category
whatnot.com/search?query=...	that search

A locale prefix such as /en-GB/ is accepted. A link that is not one of these is reported and skipped rather than guessed at.

Filters

Categories, show status, show format, buying format, condition, graded only, autographed only, minimum seller rating, Premier Shop only, seller country, and a price range. They apply in both search mode and URL mode.

Two honest notes:

• Categories are checked against the site's own live category list at the start of the run. A handle that does not exist stops the run and names the closest matches, instead of quietly returning everything. The site's category vocabulary is large and changes, so this is a free-text field rather than a frozen dropdown that would go stale.
• The price range is applied by this actor, not sent upstream. The site accepts a price filter and then returns items below the floor anyway, in every encoding its own field name suggests. Rather than ship a control that silently does nothing, the bound is enforced here against each record's own reported price, before any extra request is made for that record, so a row you do not receive never costs you anything.

Resume and recurring updates

Resume from a previous run continues one interrupted crawl from a run or dataset id, without returning or charging for what that run already collected.

Incremental mode is for scheduled monitoring. The first run returns everything as NEW; later runs return only NEW, UPDATED and REAPPEARED, so you see a newly scheduled show, a show going live, a price change or a fresh lot in a category you watch. Optionally also UNCHANGED and EXPIRED rows, both off by default because they return and bill extra rows.

Audience counters and seller follower counts are deliberately ignored when deciding what changed. They drift constantly, and counting them would mark every result changed on every run and bill the whole feed daily for nothing.

EXPIRED is only produced after a run has fully scanned the tracked search. A run stopped by the result cap, a resumed run, or a run that scanned nothing never marks anything expired.

Cost

You are charged for each result returned, plus a start charge per run. Reading each seller's full profile is opt-in and adds a small surcharge on each result that comes back carrying those extra fields, disclosed in the section caption above the toggle. Several results from the same seller share one profile read, so a run focused on a few sellers costs very little extra.

Leave the default connection selected. The other connection options are billed by traffic and cost far more.

Send results into your apps (MCP connectors)

Optionally pipe results straight into the tools you already use, over Model Context Protocol. Authorize a connector under Apify, Settings, API and Integrations, then select it in mcpConnectors. Notion gets a rich page per result and needs notionParentPageUrl; the other connectors get a best-effort write or a digest. maxNotifyListings caps how many results are written per connector per run.

This never changes the dataset. Leave mcpConnectors empty to skip it entirely.

Output

Four dataset views: Results (everything), Live shows, Items and lots and Sellers. One example row, with placeholder values:

{
  "recordType": "show",
  "showId": "00000000-0000-0000-0000-000000000000",
  "url": "https://www.whatnot.com/live/00000000-0000-0000-0000-000000000000",
  "title": "Vintage slab breaks, $1 starts",
  "status": "PLAYING",
  "isLiveNow": true,
  "startTime": "2026-09-05T00:43:25Z",
  "activeViewers": 339,
  "watchlistUsers": 113,
  "categories": ["Pokemon Cards"],
  "categorySlugs": ["pokemon_cards"],
  "tags": ["$1 Starts", "Vintage", "Graded Cards"],
  "sellerUsername": "example_seller",
  "sellerUrl": "https://www.whatnot.com/user/example_seller",
  "sellerRating": 4.9,
  "sellerReviewCount": 24451,
  "sellerFollowerCount": 180746,
  "sellerSoldCount": 160261,
  "sellerIsPremierShop": false,
  "scrapedAt": "2026-09-05T10:11:07Z"
}

Prices are returned in major units (dollars, not cents) and timestamps in UTC ISO-8601, on every record kind.

Reliability

Requests stop after five attempts rather than grinding, and a run that returns no rows tells you which of three different things happened, because each one has a different answer:

What happened	How the run ends	What to do
Nothing matched your search	Succeeds, with "No results matched"	Widen the keyword or relax the filters
Nothing could be read	Fails	Enable Apify Proxy, or try again later
The site changed its data format	Fails, naming that cause	Report the run; no keyword, filter or proxy change will help

A run that could read nothing never reports itself as an empty search, and a page that comes back in an unexpected shape is counted as a refused read rather than as the end of the results. If seller enrichment starts failing systematically it is switched off for the rest of the run, and the results you still get are not charged the surcharge.