Answers
How do I get Douyin, Xiaohongshu and WeChat data into Qwen?
Short answer
Use the QwenWork desktop app. Under Extensions → Connectors → + Add, add okdata as a custom MCP server once, with its URL and your key; after that you ask in plain words, such as "what did our rivals post on Douyin this week and what are the comments saying", and it fetches the data. Alibaba's description of the Qwen app mentions no MCP or custom connectors and says nothing about reading Douyin or Xiaohongshu links. okdata is pay per call with no monthly fee: a page of a Douyin account's videos is $0.002, a page of Xiaohongshu search $0.02, a page of an Official Account's articles $0.02. A weekly scan of three rivals across the three platforms is 30 calls, about $0.33.
"Qwen" is several products now, and only one of them can take your own data source.
| Product | Douyin, Xiaohongshu and WeChat data |
|---|---|
| Qwen app | Its App Store description lists a work assistant, scheduled tasks, an agent square of official brand agents, food delivery and government services. Nothing about adding MCP or a custom connector, nothing about reading Douyin or Xiaohongshu links |
| QwenWork desktop | Yes: Extensions → Connectors takes a custom MCP server |
| QwenWork web | Alibaba documents the custom MCP steps for the desktop client only, and its FAQ says browser automation is client-only |
| Model Studio API | The Responses API takes MCP servers over SSE only; okdata serves Streamable HTTP, so use function calling instead |
Checked against Alibaba's pages in November 2026 (Qwen app 7.6.7 on the App Store).
Connect once in QwenWork
Get a key in the console; new accounts get $0.10 of trial credit. In QwenWork desktop open Extensions → Connectors → + Add → fill in by hand. Server type Streamable HTTP / SSE, name okdata, address https://api.okdata.online/mcp, and a header Authorization: Bearer okd_…. Turn it on under Installed → Custom and start a new task; Alibaba's FAQ says connector changes only apply in a new task. Ask for your okdata balance to confirm. Full steps are on the QwenWork page.
From then on, just ask. A Douyin share link goes to Douyin video details ($0.002) and then comments ($0.002 a page); a topic goes to Douyin search or Xiaohongshu search; a Xiaohongshu link to note details ($0.02); an Official Account to its article list and full text, with no QR login. Every endpoint has its own price; you pay per call.
A weekly scan of three rivals
| Step | Calls | Cost |
|---|---|---|
| Douyin: newest page of each rival's videos | 3 | $0.006 |
| Douyin: top 2 videos each, 2 comment pages | 12 | $0.024 |
| Xiaohongshu: search each brand name, newest first, 1 page | 3 | $0.06 |
| Xiaohongshu: top 2 notes each, 1 comment page | 6 | $0.12 |
| Official Accounts: newest page of articles | 3 | $0.06 |
| Official Accounts: full text of the newest article | 3 | $0.06 |
| Weekly total | 30 | $0.33 |
QwenWork's own points are a separate bill. QwenWork also has a built-in browser connector that drives your logged-in Chrome; it is fine for a link or two, but Alibaba notes it is slower than chat, takes over the tab, and leaves captchas and QR logins to you.
Qwen app only: fetch, then upload
import os, requests
API = "https://api.okdata.online"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}
def get(path, **params):
body = requests.get(API + path, headers=H, params=params, timeout=60).json()
if body["code"] != "ok":
raise RuntimeError(f"{path}: {body['code']} {body['message']}")
return body["data"]
notes = get("/v1/xiaohongshu/notes/search", query="BrandA cat food", sort="latest",
fields="id,url,title,metrics.likes")["items"]
lines = []
for n in sorted(notes, key=lambda n: n["metrics"]["likes"] or 0, reverse=True)[:2]:
lines.append(f"{n['title']} {n['url']}")
lines += [f" - {c['text']}" for c in get("/v1/xiaohongshu/notes/comments/list", id=n["id"], fields="text")["items"]]
open("notes.txt", "w", encoding="utf-8").write("\n".join(lines))
Upload notes.txt to the Qwen app (its listing says it takes ten files at once) and ask what people complain about. Developers building on Qwen models: see Qwen-Agent and Kimi. Other assistants: Doubao, Yuanbao, WorkBuddy.
Sources
Where people asked this (read 2026-11-17)
Updated