For AI

Answers

How can a seller outside China search 1688 for suppliers and factory prices from an AI agent or a script?

Short answer

Call 1688 through an API instead of the website: okdata returns public 1688 offers to callers in any country, with no 1688 account. A keyword search costs $0.022 for 20 offers with wholesale price in yuan, minimum order and supplier; the same search on AliExpress costs $0.002, priced in US dollars. Searching 3 keywords on 1688, opening 20 offers and reading reviews for 5 of them comes to $0.726.

A Shopify seller in Texas wants to add a hand-crank coffee grinder. Before she orders samples she wants three answers: what Chinese suppliers charge wholesale, which of them are factories rather than traders, and what the same grinder sells for on AliExpress. The obvious shortcut is to tell an AI agent to browse 1688, and it often stops at the door: the browser tool in at least one widely used coding agent has refused 1688.com and aliexpress.com under a site-safety policy, read-only research included. Users reported it in June 2026, and again for 1688 in October. An API call skips the browser. The script or agent asks for the data, okdata reads the public 1688 and AliExpress pages, and the same JSON comes back wherever it runs.

Six endpoints that cover sourcing research

  • 1688 product search: offers for a keyword, in 1688's order or by sales or price, with an optional price range in yuan. Each offer has its title, price, currency, units sold and the supplier's name and member ID. With view=full you also get, under ext, the minimum order, repurchase rate, whether the supplier is marked as a factory, its years on 1688 and its location.
  • 1688 product details: one offer by ID, with its price range and tiered wholesale prices, minimum order, units sold this year, variants with stock, shipping fee and ship-from location, images and video. Most of these also sit under ext, so ask for view=full.
  • 1688 product reviews: buyers' stars, text, date and the variant they bought, 10 a page, in 1688's order or newest first.
  • 1688 supplier products: what one supplier sells, from the member ID in a search result (it starts with b2b-), with wholesale and dropshipping price, units sold, category and listing date.
  • AliExpress product search: title, price in US dollars for the ship-to country you pass (US by default), units sold and star rating, in AliExpress's order or by sales or price.
  • AliExpress product reviews: stars, text, date, the variant bought, photos and helpful votes, 20 a page. Reviews with text also carry an English translation in translation; star-only reviews have none.

Chinese in, yuan out

  • 1688 is not translated. Its titles, supplier names and reviews come back in Chinese, as they were written (only AliExpress reviews come with an English translation). Search with Chinese keywords (手摇咖啡磨豆机, not "manual coffee grinder") and let the model translate the query on the way in and the results on the way out.
  • 1688 prices are in yuan (currency: "CNY"), and min_price and max_price take yuan too. AliExpress prices are in US dollars. Convert at the day's rate and add your own freight estimate before you put the two side by side.
  • A search result carries one price per offer, the cheapest variant's. The per-unit price at each quantity break is only in product details.

The job, priced

Step Calls Cost
3 Chinese keywords on 1688, one page each 3 $0.066
Details of the 20 most promising offers 20 $0.44
Reviews for 5 of them, 2 pages each 10 $0.22
1688 subtotal 33 $0.726
The same 3 keywords in English on AliExpress 3 $0.006
AliExpress reviews for 5 competing listings 5 $0.01
First catalog page of the 2 suppliers you shortlist 2 $0.044
Whole job 43 $0.786

Prices are in US dollars. The balance is prepaid by card, Alipay or WeChat Pay, and a new key comes with $0.10 to try it.

Code

import os, requests

API = "https://api.okdata.online"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}

def call(path, **params):
    body = requests.get(API + path, headers=H, params=params, timeout=60).json()
    if body["code"] != "ok":
        raise RuntimeError(f"{body['code']}: {body['message']}")
    return body["data"]

# 1688: Chinese keywords, prices in yuan. view=full adds minimum order and supplier details under ext.
offers = call("/v1/alibaba_1688/products/search", query="手摇咖啡磨豆机", sort="sales", view="full",
              fields="id,url,title,price,currency,sold,shop.id,shop.name,ext.min_order,ext.factory,ext.location")["items"]

# AliExpress: the same product in English, prices in US dollars for the ship-to country.
retail = call("/v1/aliexpress/products/search", query="manual coffee grinder", sort="sales", country="US",
              fields="id,url,title,price,currency,sold,rating")["items"]

for o in offers[:5]:
    x = o["ext"]
    print(o["price"], o["currency"], "MOQ", x["min_order"], "factory:", x["factory"], x["location"], o["url"])
for r in retail[:5]:
    print(r["price"], r["currency"], "sold", r["sold"], "stars", r["rating"], r["url"])

Feed both lists to your model with an instruction such as "translate the 1688 titles, convert yuan at the rate I give you, and flag the offers whose price leaves room under the AliExpress price". Pass offers[0]["id"] to product details for tiered prices, and shop.id to supplier products for the rest of that supplier's range.

From Claude, Cursor or another agent

Add the okdata MCP server (https://api.okdata.online/mcp) with your key; setup for each client is on the MCP page. Then ask in plain English:

  • "Search 1688 for 手摇咖啡磨豆机 by sales, keep suppliers marked as factories, and list price in yuan, minimum order and city."
  • "Search AliExpress for 'manual coffee grinder' shipping to the US and put the 5 best sellers next to those 1688 offers."

find and describe look up endpoints and prices, read_more returns the rest of a long answer, and account shows your balance; all four are free. call runs an endpoint and is billed like the HTTP API.

Things that trip people up

  • Each page is its own call. A call that reaches 1688 is billed even when the search finds nothing or the offer has been taken down. Calls refused before they reach it, for a bad parameter or a low balance, are free.
  • Without view=full, the minimum order, factory flag, tiered prices and shipping fields are left out.
  • 1688 reviews come 10 at a time, so reading twice that many is two calls.
  • AliExpress search results do not name the store, and AliExpress product details are paused for now. Compare from search results and reviews.
  • This is research only. Nothing here messages a supplier, places an order or pays; that still happens on 1688 itself or through a sourcing agent.

Other routes

1688 runs its own open platform for developers at open.1688.com. Its documentation is in Chinese, and one developer abroad called the account setup "not exactly straight forward". Logged-in browser tools and open-source scrapers work while a 1688 session stays logged in on your machine, and break when it expires or the page changes. Sourcing agents search, negotiate and ship for a fee, which suits the buying stage; an API fits the research before it, when you want a few hundred offers in a sheet. To turn buyer reviews into product decisions, see mining competitors' reviews and product research.

Sources

Where people asked this (read 2026-11-17)

Updated

Connect once, then just ask.

Every new key comes with $0.10 to try it. One sentence connects your agent.

Start free
Connect your AI

Paste this into your agent

Read https://okdata.online/agents.md and set up okdata for me; my key is in the OKDATA_API_KEY environment variable.

Claude Code, Cursor, Codex and other agents that can open a link set themselves up from it.

MCP serverhttps://api.okdata.online/mcp