JD.com Search API

Search JD.com products by keyword and get item_id, title, price, currency, shop_name, is_jd_self, sold_count and good_rate for each one, billed in tiers by the number of rows returned, $0.002472 per tier.

JD.com Search finds JD products by keyword and returns each one's item ID, title and short title, price, review count, positive-rating share, shop, whether JD sells it directly, promotion tags and selling points. Use it to see which products come up first on JD for a search, who sells them and how buyers rate them: early product research, a first look at price bands, or comparing JD's own listings with third-party shops.

Chinese keywords usually match JD's product titles more closely. Pass item_id from a result to the JD Product details or Reviews capability to dig deeper. good_rate is the positive-rating percentage JD publishes and only means much next to the review count; review_count and sold_count are display text such as 100k+, not exact numbers.

  • Page with page or page_token, and set the page size with limit.
  • promotion and sell_point carry the promotion tag and selling point from the list.
  • Failed calls and empty results are not charged.

API health

24h health
100%
24h average time
12.38 s
  • Healthy / normal
  • Available
  • Limited
  • Mostly down
Healthy

Last 24 hours

Last 30 days

  1. 1Request

    POST /api/v1/social
    {
      "platform": "jd",
      "action": "search",
      "params": {
        "keyword": "华为手机"
      }
    }
  2. 2JSON response

    {
      "results": [
        {
          "item_id": "10237497167004",
          "title": "华为手机2026热销新机上市畅享 90 Plus 麒麟8系芯片丝滑超好用鸿蒙AI防诈6620mAh巨鲸大电池柔光护眼 星空黑【128GB】 官方标配【365天延保+蓝牙耳机+运费险】",
          "price": 1999,
          "currency": "CNY"
        },
        …
      ],
      "count": 5
    }
  3. 3Rows

    item_idtitlepricecurrency
    10237497167004华为手机2026热销新机上市畅享 90 Plus 麒麟8系芯片丝滑超好用鸿蒙AI防诈6620mAh巨鲸大电池柔光护眼 星空黑【128GB】 官方标配【365天延保+蓝牙耳机+运费险】1999CNY
    10230554159591华为mate60pro手机智能手机鸿蒙智能AI手机麒麟芯片正品国行mate60赠手机壳 Mate60Pro【雅川青】 12GB+256GB 7天试用(送手机壳88W充电器)2837CNY
  4. 4CSV

    jd-search-sample.csv item_id,url,title,short_title,price,currency,sold_count,review_count,good_rate,shop_id,shop_name,shop_url,is_jd_self,promotion,sell_point,category_ids,image_url,platform

A request with keyword goes in, and one page of products comes back; each row in results is one product, listed as a field table and a CSV file.

What data you get

The table shows a real call for 华为手机, captured October 24, 2026 at 18:28 (UTC), that returned 5 rows. Your own calls return these fields as listed in the field reference.

item_idurltitleshort_titlepricecurrencysold_countreview_countgood_rateshop_idshop_nameshop_urlis_jd_selfpromotionsell_pointcategory_idsimage_urlplatform
10237497167004https://item.jd.com/10237497167004.html华为手机2026热销新机上市畅享 90 Plus 麒麟8系芯片丝滑超好用鸿蒙AI防诈6620mAh巨鲸大电池柔光护眼 星空黑【128GB】 官方标配【365天延保+蓝牙耳机+运费险】null1999CNYnullnullnull219104371迪旺手机专营店https://mall.jd.com/index-219104371.htmlnullnullnull["9987","653","655"]https://img14.360buyimg.com/n0/jfs/t1/547105/14/3627/77818/6ad1919eFd3fb2338/0083320320125c52.pngjd
10230554159591https://item.jd.com/10230554159591.html华为mate60pro手机智能手机鸿蒙智能AI手机麒麟芯片正品国行mate60赠手机壳 Mate60Pro【雅川青】 12GB+256GB 7天试用(送手机壳88W充电器)null2837CNYnullnullnull169276555尊享数码手机拼购店https://mall.jd.com/index-169276555.htmlnullnullnull["9987","653","655"]https://img14.360buyimg.com/n0/jfs/t1/483100/15/10357/93091/6a647f40Ff1f84472/008332032085260d.jpgjd

Use cases

  • First pass at product research

    Search a category word, read price for the price band and shop_name for who sells it.

  • Watch the results for a brand word

    Read the same keyword on a schedule and compare the rank of each item_id and the promotion tags over time.

  • Compare reputation

    Read good_rate next to review_count, pick well-rated products with many reviews, then read their reviews with reviews.

  • Collect selling-point wording

    sell_point is the public selling point on the list; read it across competing products to see how they word it.

The JD Help Center and JD Union (Zeus, JOS) vs EveryInfra

Here is how the JD Help Center review rules, JD Union and EveryInfra search compare on the positive-review rate and what you need to connect.

ComparedThe JD Help Center and JD Union (Zeus, JOS)EveryInfra
Positive-review rateDefinition of a positive review: a 4-star or 5-star review a user submits counts as positive and goes into the count of positive reviews and the positive-review rate.good_rate is the positive-review percentage JD publishes; review_count is the review count as written on the list, which can be rounded.
Access you needAn appkey created in JD Union can call the API on the Union platform and can also call the Union interfaces on the Zeus platform.One EveryInfra API key in the Authorization: Bearer header. You need no Union account or appkey, and no JD login.

The JD documentation is in Chinese. The table shows a translation, checked against the JD Help Center review rules and the JD Union page (jos.jd.com/jdunion) on 2026-11-22; the pages show no update date. The Union interfaces are for promoters who read promoted products, so the table compares only the positive-review rate and access. Official sources: Review and photo rules (JD Help Center) · JD Union

Limits and gotchas

  • keyword is required; Chinese keywords usually match JD product titles more closely.
  • limit has a maximum of 30. To page, use page (an integer; without it you get page 1) or pass the response's next_page_token back unchanged as page_token. When both are present, page_token wins.
  • Paging has a fixed depth limit. On the last reachable page pagination.stop_reason is depth_limit; when the list is simply short it is end_of_list.
  • From page 2, products come from a saved list that is at most one hour old. When you jump with page, pages can share or miss a product, so de-duplicate by item_id. To read the whole list without repeats, start at page 1 and follow next_page_token.
  • sold_count and review_count are text as shown on the list (such as 100,000+), so they can be rounded, and sold_count can be missing.
  • price is a snapshot of the price visible at capture time; it may not include member, new-user, regional or targeted-coupon offers, or offers applied at checkout.

Pricing

Billing is in tiers by the number of products the page returns: each tier counts as one call at $0.002472. A tier with no rows returned is refunded, and an empty page is not charged. A jump to page N with page is billed as the full pages before it plus page N, the same as paging there one page at a time. Failed calls and empty results are not charged.

CallsCost
10,000$24.72
100,000$247.22
1,000,000$2,472.22

Totals use the list price per call.

Code samples

#
bash
curl -X POST 'https://api.everyinfra.com/api/v1/social' \
  -H "Authorization: Bearer $EVERYINFRA_API_KEY" \
  -H 'Content-Type: application/json' \
  --data-raw '{
  "platform": "jd",
  "action": "search",
  "params": {
    "keyword": "华为手机"
  }
}'

OpenAPI definition (JSON)Samples read the key from the EVERYINFRA_API_KEY environment variable; copied code never contains your key.

API reference

See the parameters, response fields, errors and billing for search, and copy the page or open it as Markdown for an AI assistant.

POST

AI prompt

#
Please help me write a data collection script with the EveryInfra JD.com Search endpoint (jd/search).

How the endpoint works:
- Request: POST https://api.everyinfra.com/api/v1/social with the header Authorization: Bearer <API key>. Read the key from the EVERYINFRA_API_KEY environment variable and never hard-code it.
- In the request body, set platform to jd and action to search, and put the parameters inside params.
- Required parameters:
  - keyword: Specifies the search query for JD.com search. This must be a search phrase, not an item ID or shop selector, and matching results do not imply exhaustive catalog coverage.
- Optional parameters:
  - page: Page number of the results, an integer; leave it out or send 1 for the first page. It runs from 1 to 150 ÷ limit, rounded up: 1–5 at limit 30, 1–30 at the default limit 5, 1–15 at limit 10.
  - page_token: Page token. Leave it out for the first page. When the response has a next_page_token, pass it back as page_token, exactly as received, to get the next page, with the other parameters unchanged.
  - limit: The most rows to return in this call, an integer from 1 to 30. The default applies when you leave it out, and a call can return fewer rows. Larger values return 422 before any charge. (default: 5)
- Returns a list of records. Main fields of each record: item_id (JD numeric product ID), url, title, short_title (Short product title as JD returns it), price, currency, sold_count (Sales figure as shown in search results…), review_count (Review count as shown in search results…), good_rate (Public positive-review percentage shown…), shop_id (JD numeric shop ID), shop_name (Public shop name), shop_url (Public JD shop page); 17 fields in total, full list at https://api.everyinfra.com/api/v1/social/catalog?platform=jd.
- Example request body:
{
  "platform": "jd",
  "action": "search",
  "params": {
    "keyword": "华为手机"
  }
}
- Price: $2.47 per 1,000 requests; failed and empty results are not charged.
- For dozens or hundreds of targets, add "mode": "async" to the body: you get a job_id at once and poll GET https://api.everyinfra.com/api/v1/jobs/{job_id} for the result.
- On 422, fix the request using the suggested parameter names and allowed values in the error; on 429 or 503, wait for Retry-After and try again.

Before writing code, ask me which targets I need, how many, which fields, and what output format I want; then confirm the plan and the rough cost with me.

Request parameters

#
NameInTypeRequiredDefaultDescription
AuthorizationheaderstringYes-Bearer followed by your API key. Create it in the console; never put it in a URL or a code repository.
platformbodystringYesjdPlatform identifier, fixed value.
actionbodystringYessearchCapability identifier, fixed value.
keywordbody.paramsstringYes-Specifies the search query for JD.com search. This must be a search phrase, not an item ID or shop selector, and matching results do not imply exhaustive catalog coverage.
pagebody.paramsintegerNo-Page number of the results, an integer; leave it out or send 1 for the first page. It runs from 1 to 150 ÷ limit, rounded up: 1–5 at limit 30, 1–30 at the default limit 5, 1–15 at limit 10. With page, the page size is the limit of this request (1–30). Every page is billed at the price of page 1, tiered by the rows on that page alone, one call per 5 rows: at limit 30 a page of 30 rows is 6 calls, at limit 5 a page is 1 call. Asking for page N directly bills N pages, the same as paging from page 1 to page N (the N − 1 pages before it as full pages, plus page N itself): limit 30 with page 5 is 30 calls in one request. Blocks for rows of page N that did not come back are refunded, and nothing is charged for the request if page N is empty (for example, the list is not that long). A page that is out of range, 0, negative, fractional, a string such as "2" or a boolean is rejected with 422 before charging. A sync call reaches the deepest page, row 150; you do not need "mode": "async". The deepest page (page 5 at limit 30) has no next_page_token and pagination.stop_reason is depth_limit; a list that is simply shorter ends with end_of_list. Page 1 is fetched fresh on every request. From page 2 on, rows are cut in order from one saved full list (up to 150 rows), and when you first ask for page 2 or later that list is at most one hour old. A page requested with page is cut from that list by its original position (page N is the limit rows starting at row (N − 1) × limit + 1); it does not know what the page 1 you got earlier looked like, so between a separately fetched page 1 and it, or between pages requested across an hour boundary, a few products can repeat or be missing: dedupe by item_id when merging. To read the whole list with no repeats, start at page 1 and follow next_page_token. pagination.page is the page number of this page, and pagination.has_more true means there is a next page: ask for page + 1, or pass next_page_token back as page_token. When both page and page_token are sent, page_token wins. A page requested with page also returns a next_page_token; the next call can use it with or without page.
page_tokenbody.paramsstringNo-Page token. Leave it out for the first page. When the response has a next_page_token, pass it back as page_token, exactly as received, to get the next page, with the other parameters unchanged. Every page keeps the page size of the limit of the request that returned the token, and a later limit has no effect. You can also skip the token and ask for a page number with page; when both are sent, page_token wins. A token is valid for 24 hours and works only for the same parameters and the same account; a token used with other parameters is rejected with 422 and not charged. Every page is billed at the price of page 1, tiered by the rows on that page alone, one call per 5 rows: at limit 30 a page of 30 rows is 6 calls, so paging to page 5 is 5 pages and 30 calls; at limit 5 each page is 1 call. Blocks for rows that did not come back are refunded, and an empty page (for example, past the end of a short list) is not charged. Paging stops at row 150 (5 pages at limit 30), and a sync call gets there: you do not need "mode": "async". The page that holds row 150 has no next_page_token and pagination.stop_reason is depth_limit; a list that is simply shorter ends with end_of_list. Page 1 is fetched fresh on every request. From page 2 on, rows are cut in order from one saved full list (up to 150 rows), and when you first ask for page 2 or later that list is at most one hour old. When you page from page 1 with page_token, every later page is cut from that same list: products already given on page 1 do not come back and nothing is skipped in between, so there are no repeats between pages. Slow paging makes no difference: the token chain keeps using the list it had when it first reached page 2. pagination.page is the page number of this page, and pagination.has_more true means there is a next page: pass next_page_token back as page_token. A page requested with page also returns a next_page_token; the next call can use it with or without page.
limitbody.paramsinteger (1–30)No5The most rows to return in this call, an integer from 1 to 30. The default applies when you leave it out, and a call can return fewer rows. Larger values return 422 before any charge.

Response example

#
Captured October 24, 2026 at 18:28 (UTC) by calling this API with the parameters in the code sample above. results and count are that call's real response with personal information (names, handles, avatars, profile links, contact details) removed; the list keeps the first 2. id, billing and quota differ on every call, so only their shape is shown.
json
{  "id": "req_…",  "platform": "jd",  "action": "search",  "results": [    {      "item_id": "10237497167004",      "url": "https://item.jd.com/10237497167004.html",      "title": "华为手机2026热销新机上市畅享 90 Plus 麒麟8系芯片丝滑超好用鸿蒙AI防诈6620mAh巨鲸大电池柔光护眼 星空黑【128GB】 官方标配【365天延保+蓝牙耳机+运费险】",      "price": 1999,      "currency": "CNY",      "shop_id": "219104371",      "shop_name": "迪旺手机专营店",      "shop_url": "https://mall.jd.com/index-219104371.html",      "category_ids": [        "9987",        "653",        "655"      ],      "image_url": "https://img14.360buyimg.com/n0/jfs/t1/547105/14/3627/77818/6ad1919eFd3fb2338/0083320320125c52.png",      "platform": "jd"    },    {      "item_id": "10230554159591",      "url": "https://item.jd.com/10230554159591.html",      "title": "华为mate60pro手机智能手机鸿蒙智能AI手机麒麟芯片正品国行mate60赠手机壳 Mate60Pro【雅川青】 12GB+256GB 7天试用(送手机壳88W充电器)",      "price": 2837,      "currency": "CNY",      "shop_id": "169276555",      "shop_name": "尊享数码手机拼购店",      "shop_url": "https://mall.jd.com/index-169276555.html",      "category_ids": [        "9987",        "653",        "655"      ],      "image_url": "https://img14.360buyimg.com/n0/jfs/t1/483100/15/10357/93091/6a647f40Ff1f84472/008332032085260d.jpg",      "platform": "jd"    }  ],  "next_page_token": …,  "pagination": { … },  "count": 5,  "billing": {    "charged": …,    "charged_usd": …,    "balance_after_usd": …,    …  },  "quota": { … }}

Response fields

#

Business data is in results. The table lists the fields a record can contain; values the platform does not show are never made up.

FieldMeaning
item_idJD numeric product ID. From search results you can pass it straight to detail, reviews, or price calls.
urlLink to the record on the platform.
titleTitle. null when the platform has no titles (for example text-only posts).
short_titleShort product title as JD returns it.
pricePrice as a number without a currency symbol; see currency.
currencyCurrency code of the price (for example CNY, USD). If price is present but currency is null, do not assume a currency.
sold_countSales figure as shown in search results (search): display text such as 100k+, not an exact unit count.
review_countReview count as shown in search results (search): display text that may be rounded, not an exact count.
good_ratePublic positive-review percentage shown by JD.
shop_idJD numeric shop ID.
shop_namePublic shop name.
shop_urlPublic JD shop page.
is_jd_selfWhether this is a JD self-operated (JD.com retail) product.
promotionPromotion label from the product list.
sell_pointPublic selling points from the product list.
category_idsJD category ID path.
image_urlImage URL: in search, product_detail, shop and skus it is the product's main (cover) image; in description_images it is the description image of that row (the full-size https image on JD's image host).
platformSource platform identifier, the same as platform in the request (for example xiaohongshu, tiktok).

Errors and billing

#
401
Missing or invalid key.
402
Insufficient balance. Top up first.
422
Invalid parameter name or value. Fix it using the suggestions in the error. No charge.
429
Per-minute request limit reached. Wait for Retry-After, then try again.
503
Could not complete. The held charge is returned to your balance. If the outcome is unknown, verify it before you retry.
200 empty
No data. No charge.

Errors

Official references

#

Used only to check objects and terms. They do not imply authorization or endorsement of EveryInfra by the platform.

  • JD.com JOS and JD Union: Product and SKU Objects ↗

    JD.com's authorized open-platform and affiliate materials describe product and SKU objects such as names, primary images, categories, displayed prices, logistics, self-operated status, and order-volume signals. This terminology does not prove that every webpage or field is available, that a caller has access, or that EveryInfra implements the official interface.

    Checked 2026-09-11

FAQ

Can I search in English?

Yes, but Chinese keywords usually match the product titles on JD more closely and return better hits. When you search for a brand and model, write it the way it tends to appear in JD product titles.

How do I get details and reviews for a result?

Each result's item_id is the numeric JD product ID. Pass it straight to the JD Product details, Reviews or Product price capability; there is no need to build a product link first.

How do I tell JD's own listings from third-party shops?

Check is_jd_self: true means JD sells the product itself. Shop details are in shop_name, shop_id and shop_url. Comparing prices and ratings within each group is fairer than mixing them.

Which fields does the JD search API return?

item_id, url, title, short_title, price, currency, sold_count, review_count, good_rate, shop_id, shop_name, shop_url, is_jd_self, promotion, sell_point, category_ids and image_url.

Should I convert item_id and shop_id to numbers?

No. item_id, shop_id and each entry of category_ids are strings, so store them as strings.

How is one page of search billed?

By the rows the page returns, in tiers: each tier is one call at $0.002472, or $2.47 per 1,000 calls. Tiers for rows not returned are refunded, and an empty page is not charged. Failed calls and empty results are not charged.

You can call the other JD.com endpoints with the same EveryInfra API key.

Google Analytics needs consent. Privacy