Custom scraper

Temu Scraper

We do not have a dedicated Temu data API. What we do have is an AI scraper that reads a temu.com page you point it at, turns it into the columns you asked for and hands them back by REST, CSV or Excel — on a schedule, priced per page.

  • temu.com
Three stages of a Temu run: paste a temu.com URL billed from the render price of EUR 0.75 per 1,000 pages; declare columns such as product_title and seller_name; and the 11 columns it returns.

How to scrape Temu

Temu publishes no read API for its catalogue. There is no key to apply for and no endpoint to call, which is why the page is the interface. Temu competes purely on landed price and reprices aggressively, which makes it the hardest kind of competitor to track and the one most often quoted internally. The way to get the data is to read the page the way a customer sees it. On a marketplace the price on the page belongs to a seller, not to the site, so a row without the seller on it is an observation you cannot act on.

How temu.com is laid out

What a run needs to know about Temu before you pick the URLs.

  • Goods ID in the URL

    Product pages follow goods.html?goods_id={id}. The goods ID is the stable key; titles change often and should not be used to match.

  • Country decides the price

    The currency and prices follow the country Temu believes the visitor is in. Pin the market for every run.

  • Local warehouse items

    Some items ship from a local warehouse with faster delivery, and search can be filtered to local warehouse only. Store the shipping origin with the price.

  • Market price and sold count

    Listings show a crossed-out market price and a sold count such as 36K+ sold. Store both, but treat the market price as Temu's own reference.

  • Lightning deals

    Lightning Deals are limited-time prices on a rotating set of items. Tag them so they are not read as the regular price.

What the AI scraper does with a temu.com link

The same three steps as any other page you point it at. Nothing about Temu is pre-built, which is exactly why it works on pages nobody wrote a connector for.

  1. Paste the link

    Give it a temu.com product or listing URL. Market, currency and assortment all come from the page itself, so there is no region setting to get wrong — the URL decides.

  2. Declare the columns

    Name each field, give it a type and write a line telling the AI what to look for. “product_title” and “seller_name” are the two most people start with. The schema is yours and you can change it between runs.

  3. Schedule it and collect

    Run it daily, weekly or on demand. Every run returns the same columns in the same order with a timestamp on each row, so the history builds itself.

Set it up once. The data keeps coming.

Save your search once and it runs by itself. Every run lands in your own Scrapewise database, and you pick how to read it.

  1. Set it up once

    Add your ASINs, keywords, places or apps to a scraper in the web app. New accounts get 5 free requests.

  2. Runs on your schedule

    Pick daily, weekly on the days you choose, every few days, or the first or last day of the month. Scheduled runs start at night, European time. Need it more often? Start runs from your own code.

  3. Saved in your database

    Each run adds dated rows, so this week sits next to last week. Rows are kept for 90 days.

  4. Use it your way

    Open the table in the web app and download the latest run as an Excel file. Ask your own AI assistant about it. Or pull the rows into your code with an API key.

Ask your AI assistant about your own data

Connect Claude Desktop or Claude Code with a read-only key. The assistant reads the rows you've already collected, at no extra cost on Scrapewise. With a full-access key it can start a run for you too.

  • Which of my ASINs lost the Buy Box this week?
  • Which competitor cut prices the most since Monday?
  • Show the keywords where I dropped out of the top 10.

What you send and what you get

You send a link and a column list. Everything else — market, currency, which products are on the page — comes from the page, so there is nothing to configure twice.

You give
  • Link to a Temu pagerequired

    Open the product or category page in your browser and copy the address. Any Temu country domain works the same way.

    https://www.temu.com/
  • Your column listrequired
    product_title, seller_name, price, currency
  • How often it runsoptional
    Daily at 06:00 UTC
You get

One row per listing, 11 columns each

  • product_title
  • seller_name
  • price
  • currency
  • shipping_cost
  • availability
  • rating
  • review_count
  • is_featured_offer
  • product_url
  • checked_at

REST API, CSV or Excel. The column set stays the same between runs, so downstream jobs do not re-map fields.

The columns a Temu run comes back with

We are not going to print a table of invented marketplace prices and call it real output. Temu answers an ordinary request with Cloudflare rather than a product page, so we have no run of our own to quote from. So what is published here is the column list you declare before the run, which is the same list every run comes back with.

product_titleListing title as the seller wrote it
seller_nameWhich seller this price belongs to
pricePrice this seller is charging today
currencyCurrency the page quoted
shipping_costDelivery charge shown for this listing
availabilityWhether the listing is in stock
ratingAverage rating, where the page shows one
review_countHow many reviews back that rating
is_featured_offerWhether this listing is the one the page defaults to
product_urlLink back to the page the row came from
checked_atWhen this run collected the page (UTC)

Every column here is one you name yourself in the schema. If a field is not on the page on the day of the run, it comes back empty rather than guessed.

What a Temu page costs

Pay-as-you-go from your wallet. No plan, no monthly fee, no seat count. The price shown is the starting price: €0.75 per 1,000 pages with a browser render. If a run also needs a residential proxy it costs €3.75 per 1,000 pages.

€0.75per 1,000 pagespay-as-you-go, no plan

Measured on 3 October 2026, not assumed. A plain request for a temu.com page returns a document rather than a refusal, but Cloudflare is in front of it, and the parts of the page that carry price are filled in by JavaScript after load, so a browser is what makes the run reliable.

Worked example

500 Temu pages checked once a day for a month is 15,000 pages.

about EUR 11.25 for the month at the render price, about EUR 56.25 if the pages also need a residential proxy

You are charged for what a run actually used, not what it was expected to need, and your balance never expires.

See the full price list →

Typical fields the AI scraper extracts from a Temu page

Treat this as a starting point rather than a fixed schema. What comes back is whatever the page actually shows on the day of the run.

Who owns the price

The distinction that makes marketplace data usable: the same product page can carry a dozen prices from a dozen sellers.

  • Seller name as the listing shows it
  • Whether this is the offer the page defaults to
  • Seller rating and feedback count, where shown
  • Number of other offers on the same product
  • Fulfilment or delivery promise shown next to the price

The landed price

A headline price that excludes delivery is not comparable across sellers, and on a marketplace it usually does.

  • Price shown on the listing
  • Delivery charge shown for this seller
  • Any strike-through or was-price the page prints
  • Promotion or voucher wording
  • Currency and market the price belongs to

Listing identity

Marketplaces issue their own product and listing identifiers, which is the most reliable thing on the page to key a comparison on.

  • Listing title
  • Site's own product or offer identifier
  • Brand, where the listing declares one
  • GTIN or EAN, where the listing publishes one
  • Category breadcrumb
  • Product image URLs

The column list is one you write: name each field, give it a type and a line telling the AI what to look for. The schema belongs to your run, not to us. How Custom Schema works →

What it cannot give you If a field is not visible on the page, the scraper cannot invent it. Identifiers such as EAN or GTIN only come back when Temu publishes them.

What people use Temu data for

01

Watching who undercuts you on your own listings

If you sell here, the useful question is not what the page says but which seller is currently winning the default offer and by how much. That is a seller column next to a price column, read daily.

02

Unauthorised-reseller detection

Brands usually find out their product is being sold below their own floor by accident. A daily sweep of your own products with the seller name attached turns that into a report instead of a surprise.

03

Assortment and gap research

What a category actually carries, at what price points, from how many sellers. A few weeks of scraped listing data answers that before you commit to a range.

04

Price floor enforcement

You already know your minimum advertised price. What you do not have is a list of listings below it, with the seller and the date, which is the evidence a channel conversation needs.

What makes Temu harder than an ordinary storefront

Here is what the scraper is actually up against on temu.com, measured rather than assumed.

Access

  • Plain HTTP requests are answered, but the answer is the shell of the page rather than the priced version of it
  • Named protection on the response: Cloudflare
  • No usable sitemap directive in robots.txt, so the URL list comes from your own category pages

One page, many prices

  • The same product page carries several sellers' offers, so a single price column is ambiguous without a seller column
  • Which offer the page shows first changes through the day, independently of any price changing
  • Delivery cost is often where the real competition happens and is quoted separately from the price
  • Listing pages are lazily loaded and paginated, so a category sweep is a list of URLs rather than one request

What the page publishes

  • No product-level structured data is published, so every field is read off the rendered page
  • Prices are personalised per session and per market, so a run has to pin the market explicitly or it will record a price no customer was ever shown.
  • Price is one of the last things the page settles on, so a read taken too early records the placeholder rather than the number

What to watch on temu.com

The one thing specific to this site. The general limits — public pages only, no field the page does not show, your URLs rather than a full crawl — are in the custom scrapers FAQ.

  • Temu has one specific trap

    Prices are personalised per session and per market, so a run has to pin the market explicitly or it will record a price no customer was ever shown.

FAQ

Temu scraping — questions

What people ask before pointing the scraper at temu.com.

Temu publishes no read API for its catalogue. There is no key to apply for and no endpoint to call, which is why the page is the interface.

Questions that apply to every site — scheduling, exports, what it will not do — are answered in the custom scrapers FAQ.

Ready to pull Temu data into your stack?

Start free — or talk to our team about your exact fields, refresh cadence and volume.