Store21vek.by

21vek scraper API for AI agents and code

datawebot returns 21vek search results as structured JSON, with the same fields for every shop: title, price, original price, availability, rating, image and link. It runs from your AI agent over MCP or from code over REST on 21vek.by in Belarus, at 5 credits per search page, and failed requests are free.

Fields you get

Every 21vek product comes back in these fields; a field is null when the page doesn't show it. Each endpoint below also lists its raw record schema.

Field	Type	Description
`product_id`	string or null	Opaque id. Pass it to get_product (or capability "product") to read the product again.
`title`	string	Product title as the shop shows it.
`url`	string or null	The product page on the shop.
`shop`	string or null	The shop’s hostname, without www.
`price`	object or null	amount (number), currency (ISO 4217, or null when the shop doesn’t say) and text (the shop’s own wording when it is the same amount, else null).
`original_price`	object or null	The price before a discount, in the same shape as price.
`rating`	number or null	Average rating, on the shop’s own scale.
`review_count`	integer or null	Number of reviews or ratings.
`image_url`	string or null	Main product image.
`availability`	string or null	in_stock, out_of_stock, or the shop’s own stock wording.
`sponsored`	boolean or null	true for an ad the shop marks as sponsored; null when the page doesn’t say.

How to run it

From your agent, after connecting it once with the setup prompt (agent setup), ask in plain words:

Agent prompt

Use datawebot to extract https://www.21vek.by/search/?term=%D0%BD%D0%BE%D1%83%D1%82%D0%B1%D1%83%D0%BA and give me each product's title, price and link as JSON.

From code, send one request with your API key. The answer has the same shape for every shop; if it says "status": "working", read it again with GET /v1/requests/{id}.

POST /v1/requests

curl https://api.datawebot.com/v1/requests \
  -H "Authorization: Bearer $DATAWEBOT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"capability":"endpoint","endpoint":"vek21.search_items","url":"https://www.21vek.by/search/?term=%D0%BD%D0%BE%D1%83%D1%82%D0%B1%D1%83%D0%BA","format":"structured"}'

Search

21vek.by Belarus (www.21vek.by) keyword search: the first 1-5 main-grid cards of the first results page in grid order, with the decimal product code, product URL, title, the displayed card price (BYN only when the card states it), the source regular price, labels and raw source status. Search only: no product pages, pagination, filters or sorting.

ID: vek21.search_items
Version: 1
Price: 5 credits / page
State: Available
Typical time: About 30 s (estimate)

What it does

Reads one search results page (or up to the page limit) and returns one record per product card, in page order. Works on www.21vek.by; other hosts are rejected before anything runs.

Input

One page URL per request, shaped like https://www.21vek.by/search/?term=%D0%BD%D0%BE%D1%83%D1%82%D0%B1%D1%83%D0%BA, plus these bounded parameters.

Parameter	Type	Allowed	Default	Set in /v1/requests
`max_products`	integer	1–5	5	`POST /v1/jobs` only

Price

5 credits / page. Failed requests are free; a page that loaded but had nothing on it still costs credits. Set max_credits to refuse anything dearer before it runs.

Output

The default, "format": "structured". Every shop answers in the same fields, mapped from the page to the product_summary@1 schema. Fields are null when the page doesn't show them, and a record that can't be mapped is read best-effort from its raw fields. Products are in products, up to 50 a page; follow next_cursor with GET /v1/requests/{id} (MCP: get_results).

Field	Type	Description
`product_id`	string or null	Opaque id. Pass it to get_product (or capability "product") to read the product again.
`title`	string	Product title as the shop shows it.
`url`	string or null	The product page on the shop.
`shop`	string or null	The shop’s hostname, without www.
`price`	object or null	amount (number), currency (ISO 4217, or null when the shop doesn’t say) and text (the shop’s own wording when it is the same amount, else null).
`original_price`	object or null	The price before a discount, in the same shape as price.
`rating`	number or null	Average rating, on the shop’s own scale.
`review_count`	integer or null	Number of reviews or ratings.
`image_url`	string or null	Main product image.
`availability`	string or null	in_stock, out_of_stock, or the shop’s own stock wording.
`sponsored`	boolean or null	true for an ad the shop marks as sponsored; null when the page doesn’t say.

Answer shape (types, not real data)

{
  "request_id": "string",
  "status": "done",
  "products": [
    {
      "product_id": "string or null",
      "title": "string",
      "url": "string or null",
      "shop": "string or null",
      "price": {
        "amount": "number",
        "currency": "string or null",
        "text": "string or null"
      },
      "original_price": {
        "amount": "number",
        "currency": "string or null",
        "text": "string or null"
      },
      "rating": "number or null",
      "review_count": "integer or null",
      "image_url": "string or null",
      "availability": "string or null",
      "sponsored": "boolean or null"
    }
  ],
  "next_cursor": "string or null",
  "meta": {
    "as_of": "date-time",
    "country": "string",
    "credits_used": "integer",
    "credits_left": "integer"
  }
}

Over the Jobs API, GET /v1/jobs/{id}/results?format=structured returns the full product_summary@1 records.

With "format": "raw", the answer keeps the structured fields and adds raw: this endpoint's own records, unchanged, one per item, in the schema below. It is the data the endpoint extracted from the page, not the page's HTML. Same price as structured.

Field	Type	Description
`rank`	integer required	1-based main-grid position on page 1
`product_id`	string required	Literal decimal 21vek product code
`url`	string (uri) required	Product URL without the per-search searchId parameter
`title`	string required	—
`price`	string required	Displayed card price (salePrice, or price when salePrice is null), cross-checked with the page state; not installments or checkout
`currency`	"BYN" \| null required	BYN only when the card states it; otherwise null (not inferred)
`regular_price`	string or null required	Source state price field, an observation
`price_meaning`	string required	—
`labels`	array required	Card labels (discounts, installments) as source text
`source_status`	string or null required	Raw source status (e.g. in, request); not normalized availability
`availability`	null required	—
`variant_id`	null required	—
`seller_id`	null required	—
`seller_name`	null required	—
`offer_id`	null required	—
`shipping`	null required	—
`vat`	null required	—

One raw record (types, not real data)

{
  "rank": "integer",
  "product_id": "string",
  "url": "string (uri)",
  "title": "string",
  "price": "string",
  "currency": "BYN | null",
  "regular_price": "string or null",
  "price_meaning": "string",
  "labels": "array",
  "source_status": "string or null",
  "availability": null,
  "variant_id": null,
  "seller_id": null,
  "seller_name": null,
  "offer_id": null,
  "shipping": null,
  "vat": null
}

Raw needs this endpoint: a search with format: "raw" doesn't fall back to shopping search, and answers raw_unavailable where no endpoint can run. Over the Jobs API, results are raw by default; format=both gives each raw record with its structured form.

Request

Run exactly this endpoint with capability: "endpoint"; there is no fallback, and a failed run is free. Set format to "raw" for the raw records too. For a keyword instead of a URL, search_products with this shop as site builds the search URL, tries the shop's endpoint first and then shopping search.

POST /v1/requests

curl https://api.datawebot.com/v1/requests \
  -H "Authorization: Bearer $DATAWEBOT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"capability":"endpoint","endpoint":"vek21.search_items","url":"https://www.21vek.by/search/?term=%D0%BD%D0%BE%D1%83%D1%82%D0%B1%D1%83%D0%BA","format":"structured"}'

MCP submit_job

submit_job {
  "endpoint": "vek21.search_items",
  "url": "https://www.21vek.by/search/?term=%D0%BD%D0%BE%D1%83%D1%82%D0%B1%D1%83%D0%BA",
  "format": "structured"
}

If the answer says "status": "working", read it again with GET /v1/requests/{id} (MCP: get_results) after check_after_seconds.

Limits

One page URL per request, on 21vek.by only; other hosts are rejected before anything runs.
Prices and stock are read when the request runs (meta.as_of); datawebot keeps no price history, so schedule and store repeat reads yourself.
Fields the page doesn't show come back null; nothing is filled in or guessed.
Up to 30 requests a minute per account.

Questions about the 21vek scraper API

Does 21vek have an official API?

datawebot does not use one. It reads 21vek's pages and returns them as structured JSON, so you need no 21vek developer account or partner key. Whether 21vek offers its own API, and on what terms, is for 21vek to say.

Can I scrape 21vek with Python?

Yes. Send the POST /v1/requests call shown on this page from Python with any HTTP client, such as requests or httpx: the same JSON body, with your API key in the Authorization header. The products come back in the answer's products field (product for a product page).

Is it legal to scrape 21vek?

It depends on where you are, 21vek's terms and what you do with the data. You stay responsible for how you use what datawebot returns. This is not legal advice; read our guide to robots.txt and legal considerations, and ask a lawyer about your case.

What does a 21vek scraper API call cost?

5 credits per search page. Failed requests are free; a page that loaded but had nothing on it still costs credits. Every account starts with 200 free credits, and credit packs start at 10,000 credits for $10, with no subscription.

Start with 200 free credits

Connect your agent with one setup prompt, or create an API key, and run 21vek requests right away. No card and no subscription.

Connect your agent See pricing