ThredUp Real-Time Data

Live ThredUp resale intelligence — search listings by keyword for price, brand, size, condition, discount and demand signals, plus full details, seller shop inventory and completed-listing comps. Luxury finds are tagged, never filtered, so 1,000 records stay 1,000. Every record streams to your dataset the moment it is collected, in clean, structured JSON.

Free trial: on a free Apify plan, runs are capped at 2 results so you can verify the data quality. Upgrade to a paid plan for unlimited exports.

---

What you can do with it

ThredUp is one of the deepest secondhand catalogs online — millions of pre-owned and new-with-tags items across brands from Carhartt and Patagonia to Farm Rio and Free People. This actor turns that catalog into structured resale intelligence you can act on.

You want to…	Use this
Find undervalued inventory (vintage single-stitch, designer denim, gorpcore, Y2K)	Listing Search with a minimum discount filter
Catch fresh drops before other resellers	Listing Search with Listed within (days) + webhooks
Price your own inventory against the live market	Listing Search + Sold Comps
Track a specific listing's full spec sheet	Listing Details
Monitor a partner shop's entire inventory	Seller Shop Listings
Vet a shop before you buy or partner	Seller Profile
Collect any ThredUp page you already have a link for	Scrape By URL
Get Slack/Discord deal alerts	Webhook URL
Ask an AI agent for comps	Apify MCP

Who it's for: vintage & thrift resellers (cross-listing on Poshmark, eBay, Depop, Mercari, Grailed), consignment shops, resale arbitrageurs, pricing analysts, fashion researchers, and AI agents that need live resale data.

---

Quick start (10 results in seconds)

1. Keep Listing Search enabled (it is on by default).
2. Keywords are already prefilled: carhartt detroit jacket, vintage 90s levis 501, patagonia fleece.
3. Click Start.

You get one row per listing — title, brand, size, condition, prices, discount, save count, images, and more — as it lands.

---

Features

Feature	Checkbox	What it returns
Listing Search	enableListingSearch (default on)	One row per matching listing, streamed live, with keyword, rank, and total market depth.
Listing Details	enableListingDetails	Full records for specific item numbers or product URLs: description, fabric, care, features, pattern, measurements, MPN, and full image gallery.
Seller Shop Listings	enableClosetListings	One row per live item in a seller shop. (ThredUp is shop-based rather than closet-based, so this uses shop IDs — featureType stays closet_listings for cross-platform compatibility.)
Seller Profile	enableSellerProfile	One row per shop: shop type, live inventory depth, and completed-listing depth.
Sold Comps	enableSoldHistory	One row per completed listing where the marketplace publishes it — sold price, retail baseline, and discount from retail.
Scrape By URL	enableScrapeByUrl	Paste any ThredUp search, department, brand, shop, or product URL; the page type is detected automatically.

Enrichment happens in place — no duplicate rows

Enrich with full listing details (searchFetchFullDetails) merges the full record into the same listing_search row and sets detailsFetched: true. You never get two rows for one listing, and partial items are never dropped — every discovered item reaches your dataset.

Grading labels tag rows — they never remove them

Luxury brand tier labels (labelLuxuryBrands) and the discount target (labelMinDiscountPercent) grade each listing instead of filtering it: a 1,000-row request stays 1,000 records, with brand_tier, is_luxury_brand, and meets_discount_target filled in for the segment you care about. Nothing is filtered for missing a tier, so output volume — and therefore per-1,000-record pricing — stays predictable. Details in Grading labels.

---

Input reference

Run settings

Field	Type	Default	Notes
country	US | UK	US	Market region; also drives the recommended proxy country.

Listing Search

Field	Type	Default	Notes
enableListingSearch	boolean	true	Live keyword search.
searchKeywords	string list	3 prefilled thrift queries	One keyword per entry, processed in parallel.
searchMaxResults	integer 1–1000	10	Per keyword. Total rows = keywords × this number, bounded by the run limit.
searchSort	enum	relevance	relevance, newest_first, price_low_high, price_high_low, marked_down_at_desc, recommended — these are the marketplace's own sort options.
searchDepartment	string	–	e.g. women, men, kids, home.
searchBrand	string	–	e.g. Carhartt.
searchCondition	string	–	As the marketplace labels it: excellent, good, fair.
searchMinPrice / searchMaxPrice	number	–	Price band in USD.
searchListedWithinDays	integer 1–365	–	Only recently listed inventory — ideal for fresh-deal alerts.
searchFetchFullDetails	boolean	false	Merge full details into the same row.

Grading labels — tag listings, never filter them

Field	Type	Default	Notes
labelLuxuryBrands	boolean	false	Tag each listing with the marketplace's luxury brand grouping. Nothing is removed.
labelMinDiscountPercent	integer 1–95	–	Savings-off-retail target to grade against; sets meets_discount_target on every row.

Labels grade what you collected instead of deciding what you collect. Every discovered listing is written to the dataset in full, matching or not matching the label, so a labeled run returns exactly as many rows as an unlabeled one — the record count stays priceable, and there are no duplicate rows. Filter on the label fields afterwards, in your own tooling, to isolate the segment you care about.

Cost: a little extra time per keyword to resolve the tags — bounded by how many rows the run can export, so a small run stays quick and a capped run never pays for tags it will not write. The row count never changes. If tags cannot be resolved for a keyword, the run continues and the label fields are simply left blank — listings are still saved.

Run limits

Field	Type	Default	Notes
maxItems	integer 0–1,000,000	0	Total rows for the whole run, across every enabled feature. 0 = no run limit.

Narrowing filters — they define the search, and the run tells you the pool size first

Brand, department, condition, price band, and freshness narrow which inventory you are shopping. Every listing that matches is still written in full — nothing is dropped for being incomplete — but a narrower pool caps how many rows can exist. So the run measures that pool before it collects anything:

Projected output: 1000 record(s) — 1 keyword(s) x up to 1000, bounded by the run limit and by how many listings actually exist
"jacket": 10,001 listing(s) available, collecting 1000
Narrowing active (Brand / designer) — these reduce how many listings exist, so your output is capped by the availability above

If the pool is smaller than your target, you are told in plain numbers ("jacket": only 40 listing(s) match the current filters (you asked for 1000)) and the run collects everything that matches rather than silently under-delivering. The same numbers land in the run OUTPUT (projectedRecords, keywordPools), so a caller can price the run before paying for it.

Measured on the single keyword jacket against the live marketplace, for orientation: no filters 10,001 · listed within 1 day ≈1,700 · brand or department scoped — varies by brand. Anything that cannot be planned is reported up front, never discovered at the end.

Bulk price research: how to reliably land ~1000 rows

1. Leave all narrowing filters empty — they are for deal hunting, not for bulk comps.
2. Give one or two broad keywords (jacket, jeans, sweater) and set Max results per keyword to the number you want.
3. Optionally set Run limit so a multi-keyword run cannot overshoot.

Measured locally with filters off: 300 rows in 7.1s, 1000 rows in 20.3s, zero duplicates. Paging is offset-stable, and repeated listings across page boundaries are skipped, so a request for 1000 unique rows delivers 1000 unique rows rather than “1000 attempted”. If the pool genuinely runs out first, the log says so instead of silently returning less.

Targeting a luxury segment without breaking that math

Luxury inventory is roughly one listing in ten for a broad keyword, and the share changes per keyword. So the actor reports it instead of guessing:

"jacket": 968 of 10,001 matching listing(s) are in the luxury group — they are tagged, not filtered out
"jacket": 412 of 1000 saved listing(s) carry the luxury tag

With Luxury brand tier labels on, a 1000-row run still returns 1000 rows; the share line tells you exactly how many carry the tag, and OUTPUT.luxuryTagged / OUTPUT.luxuryShare report the same thing to a calling system. To land ~1000 luxury rows, divide by the reported share — e.g. a 9.7% share means asking for ≈10,300 rows — which is a number you can read before the run finishes instead of a number you have to guess.

Filters this marketplace does not expose

ThredUp does not offer a per-size filter on its search surface, and this actor does not fake one — size is returned on every row, so filter the dataset instead. Condition is a free-text filter rather than a fixed dropdown, matching how the marketplace itself accepts it. Everything else above is a real, working filter.

Listing Details

Field	Type	Notes
listingIds	string list	ThredUp item numbers, e.g. 1500810136.
listingUrls	string list	Full product URLs.

Seller shops (shared by three features)

Field	Type	Default	Notes
sellerIds	string list	–	Seller shop IDs. Shared by Seller Shop Listings, Seller Profile, and Sold Comps.
sellerMaxListings	integer 1–200	30	Live listings per shop.
soldMaxItems	integer 1–200	30	Completed listings per shop.

Scrape By URL

Field	Type	Notes
scrapeUrls	string list	Any ThredUp URL. URLs from other sites are rejected with a clear error.

Supported URL shapes

https://www.thredup.com/women?search_text=carhartt%20detroit%20jacket
https://www.thredup.com/women/carhartt
https://www.thredup.com/product/women-carhartt-jacket/1500810136
https://www.thredup.com/shop/<shop-id>

Search URLs are read exactly as the marketplace writes them — including department_tags, brand_name_tags, price[min], price[max], condition, clearance, luxe_brand, user_promotion_discount_percent, and listed_days — so any link you copy from a ThredUp browsing session works as-is.

Alerts

Field	Type	Default	Notes
webhookUrl	string	–	Every record is POSTed here right after it is saved. Delivery never slows collection.
webhookFormat	json | slack	json	json = full record; slack = ready-to-read message.
proxyConfiguration	object	Apify Residential	Residential proxy is recommended for stable runs and correct regional pricing.

---

Output reference

One dataset row per item, streamed as it is collected. featureType tells you which feature produced the row: listing_search, listing_details, closet_listings, seller_profile, sold_history, or scrape_by_url.

Shared core (every listing-like row)

Field	Description
featureType	Which feature produced this row.
scrapedAt	ISO-8601 write time.
url	Source URL.
item_id	ThredUp item number — reuse it with Listing Details.
item_url	Direct product link.
title, description	Listing copy.
main_image_url, additional_image_urls	Full gallery.
status	available, sold, …
current_price	Live selling price (USD).
original_price	The listing's stated original price.
retail_price	List price / MSRP.
savings_amount	Retail minus current price.
discount_percentage	Percent off retail as published. Negative means priced above retail.
currency	USD.
brand, brand_id	Brand identity.
size, size_scale	Size plus the size system (ALPHA, NUMERIC).
condition, quality_code, quality_type, condition_description	Condition as published.
category, category_tags, department, department_tags	Taxonomy.
color, colors, style_tags, material	Attributes.
likes_count	Shopper saves — a live demand signal.
is_sold	Completion flag.
seller_id, seller_type, seller_on_vacation, seller_covered_shipping, seller_shipping_cost	Seller identity when the listing belongs to a marketplace shop.
shipping_cost, free_shipping	Shipping economics.
detailsFetched	true once full details were merged in.

Feature-specific fields

Field group	Fields	Where
Search context	search_keyword, position, total_results_available	listing_search
Detail enrichment	fabric, care_instructions, features, pattern, measurements, measurements_display, mpn, mpn_title, size_detailed, photo_count	listing_details, and enriched search rows
Shop inventory	seller_shop_id, shop_total_listings	closet_listings
Shop profile	seller_type, shop_total_listings, shop_total_sold, average_rating, ratings_count, is_marketplace_shop	seller_profile
Comps	sold_price, discount_from_retail_percentage, shop_total_sold	sold_history
URL runs	pageType, item_id	scrape_by_url
Grading labels	brand_tier, is_luxury_brand, meets_discount_target	listing_search (present only when the matching label is switched on)

Runs also write an OUTPUT summary: totalPushed, spendingLimitReached, a paywall object describing the tier and whether a cap was applied, the volume projection (projectedRecords, keywordPools with each keyword's available and luxuryAvailable), the narrowing and label controls that were active (narrowingActive, labelsActive), and the label results (luxuryTagged, luxuryShare, discountTargets).

Price field semantics: current_price is the live price a shopper pays, original_price is the item's stated original price, and retail_price is the list price. discount_percentage is computed against retail_price, so it is directly comparable with comps from other marketplaces.

---

Deal alerts with webhooks

Add a Slack (or Discord) incoming-webhook URL as webhookUrl, set webhookFormat to slack, and every fresh listing shows up as a message the moment it is collected — with title, brand, price, discount, and a direct link. Webhook delivery is fire-and-forget, so alerting never slows the run, and a failed delivery never costs you a dataset row.

Example Slack card:

🛍️ Carhartt Detroit Jacket — $41.99 (48% off $80 retail)
Brand: Carhartt · Size: M · Condition: Good
https://www.thredup.com/product/...

For JSON consumers you get the full record — ideal for pricing bots, cross-listing tools, and your own dashboards.

---

AI agents & MCP

Connect the Apify MCP server and ask questions in plain language:

• "What is the average price of a 90s Carhartt J97 jacket right now on ThredUp?"
• "Find Patagonia fleece listings under $40 that are at least 50% off retail."
• "How many Carhartt jackets are listed and how deep is that market?"

Because every row carries brand, size, condition, current_price, retail_price, discount_percentage, and likes_count, an agent can compute price bands, discount distributions, and demand signals without extra work. Prefer the Dataset views (overview, search, details, closet_listings, seller_profile, sold_history, scrape_by_url) for clean, focused tables.

---

Pricing, free tier, and limits

• Billing: pay-per-event, charged per result written (result). You only pay for rows you actually receive.
• Spending limits: the actor respects your Apify spending limit. When it is reached the run stops cleanly, writes its summary, and exits gracefully — no error, no partial garbage.
• Free plan: capped at 2 results per run, then the run stops with a clear upgrade message. A hard block mode is available to account owners for validation-only runs.
• Memory & runtime: 512 MB default with a 10,000-second timeout; memory stays flat during 10,000+ item runs because rows stream to the dataset instead of buffering.

---

FAQ

How fresh is the data? Every run reads live marketplace inventory at run time — there is no stale cache.

Do I need a proxy? On Apify, residential proxy is configured by default and recommended. Locally, set your proxy in .env (see .env.example).

Why did a keyword return no rows? The narrowing filters were too tight (a rare brand combined with a narrow price band and a short freshness window). Loosen one filter at a time. Note that grading labels never cause this — turning on luxury tagging cannot reduce your row count.

Why is some optional field empty? Some listings genuinely publish no fabric, measurements, or condition notes. The actor never invents values — empty means the marketplace did not publish it.

Can I get one row per listing with everything in it? Yes: enable Enrich with full listing details and you get the search row plus the full record in a single row (detailsFetched: true).

Where do seller shop IDs come from? From ThredUp partner shop links you already have. The actor never guesses shop IDs: if a value does not resolve to a shop, that feature logs it clearly and the rest of the run continues normally.

How do I keep dataset size and cost predictable? Use searchMaxResults, sellerMaxListings, soldMaxItems, and your Apify spending limit. Labels never change the row count, narrowing filters report their pool size up front in the log and in OUTPUT.projectedRecords, and rows stream as they are found — so a spending cutoff never loses what you already paid for.

---

Local development

npm install
cp .env.example .env          # add your proxy settings
cp local.input.example.json local.input.json
npm run start:local           # streams to output/local_results.jsonl

npm run typecheck
npm run build

---

License

ISC.