Your own scraper or a data API? Cost, upkeep and account risk (2026)
It depends on how much and who will keep it running: for a little now and then, with someone to maintain it, write a scraper; on a schedule, in quantity, across platforms, or for an AI, use a data API. A scraper means keeping accounts, handling signatures and risk controls and repairing it after each change on the platform; a data API takes one key and is paid by the call. With DataXAPI a page of Xiaohongshu note search is $0.0200, with no account of yours. DataXAPI's endpoints here are from $0.0020 a call, and a failed call is not charged; what others do was checked on 2026-11-20, with sources below.
Updated 2026-11-20
How do they differ?
Side by side on what each gets, the account it needs, cost, upkeep and risk.
| Aspect | Your own scraper | A data API (DataXAPI) |
|---|---|---|
| What you get | What the site and app show, as far as you can keep it working | The fields each endpoint's page lists, each with a recorded real response |
| Account and sign-in | Mostly platform accounts and cookies, and proxies for much | A key; no platform account |
| Cost | A developer's time to build and keep it, and accounts and proxies | Per call, each endpoint's price on its page; a failed call is not charged |
| Upkeep | Repairs whenever a platform changes its signatures or endpoints, and a parser for each platform | None on your side; one shape for every platform, so another needs no new parser |
| Risk | Going fast gets limited, at worst banned | No account of yours, so none is at risk |
What does each platform put in a scraper's way?
The table is from open-source projects' code and issue reports, not from the platforms; how to get past them is not covered here.
| Platform | What you handle | What stops you | DataXAPI's endpoint |
|---|---|---|---|
| Xiaohongshu | The signing headers X-S, X-T, x-S-Common, an account's cookie | HTTP 461, code 300011 (abnormal account detected) | xiaohongshu.notes.search, $0.0200 a page |
| Douyin | The a_bogus signature, cookies and parameters such as msToken and ttwid | No data without the right signature and parameters | douyin.videos.search, $0.0200 a page |
| Bilibili | The wbi signature, the session and the IP | -352, and HTTP 412 somewhat faster | bilibili.users.videos.list, $0.0020 a page |
The mobile site's endpoints at m.weibo.cn, cookies | 432 Client Error, usually a limit for a while | weibo.users.posts.list, $0.0020 a page | |
| Official Accounts | Exporting through your own account's back end | The open-source wechat-article-exporter stopped on 2026-07-30, saying the WeChat interface it relied on was closed | wechat_mp.accounts.articles.list, $0.0200 a page |
Which one?
A little, once, with someone on the team to keep it running, and something no endpoint has: write your own. On a schedule, in quantity, across platforms, or into a spreadsheet, BI or an AI, without putting your own accounts at risk: a data API; check its fields on the free trial before you decide.
What a schedule comes to
10 words searched 5 pages deep on Xiaohongshu every day are 1,500 calls a month, $30.00. 100 Bilibili uploaders checked weekly come to about 400 calls a month, $0.80. 10 words searched 5 pages deep on Weibo every day, $3.00 a month.
What does a real call look like?
Below is the request of xiaohongshu.notes.search and a response recorded for real, with the day it was recorded under it. Each note's ID is in id, the author in author, and likes, comments, shares and saves in metrics; for the next page pass page.next_cursor back as cursor. A scraper's search, paging and parsing are one call here. The code reads the key from the environment variable DATAXAPI_API_KEY; with your own key it runs as it is.
curl -G "https://api.dataxapi.com/v1/xiaohongshu/notes/search" \ --data-urlencode "query=咖啡" \ -H "Authorization: Bearer $DATAXAPI_API_KEY"
# pip install dataxapi from dataxapi import DataXAPI dx = DataXAPI() # reads DATAXAPI_API_KEY result = dx.xiaohongshu.notes.search(query="咖啡") print(result["data"]["items"])
// npm i @dataxapi/sdk import { DataXAPI } from "@dataxapi/sdk"; const dx = new DataXAPI(); // reads DATAXAPI_API_KEY const result = await dx.xiaohongshu.notes.search({ query: "咖啡" }); console.log(result.data.items);
dataxapi call xiaohongshu.notes.search 'query=咖啡'{
"code": "ok",
"request_id": "d2403177-e4f6-4d04-99eb-afce599b4c31",
"data": {
"items": [
{
"id": "67d5e2be00000000d803d22d",
"url": "https://www.xiaohongshu.com/explore/67d5e2be00000000d803d22d",
"type": "image",
"title": "实测一个月咖啡,结论有点意外",
"text": "把咖啡从头到尾讲清楚,看完少走弯路",
"published_at": "2026-05-17T05:25:25.000Z",
"author": {
"id": "6130783200000000883b2a1e",
"username": "coffee_lab_sh",
"name": "山野咖啡馆",
"avatar_url": "https://sns-avatar-qc.xhscdn.com/avatar/610f17040000000049dd4887.jpg"
},
"metrics": {
"likes": 48,
"comments": 21,
"shares": 5,
"favorites": 10
},
"images": [
"https://sns-webpic-qc.xhscdn.com/20261010/2d7c7c32da67d3c3d839!nd_dft_wlteh_webp_3"
]
},
// … 14 more
]
},
"page": {
"next_cursor": "c1.Ip12-f4dS4HmSDpdPex7WuLPBr1zJAbdhL_Wj8K0",
"cursor_expires_at": "2026-10-12T19:43:26.000Z",
"has_more": true,
"returned": 15
},
"warnings": [],
"meta": {
"capability": "xiaohongshu.notes.search",
"result_status": "complete",
"fetched_at": "2026-10-12T18:43:26.000Z",
"cached": false,
"replayed": false,
"unavailable": {
"lang": "capability",
"metrics.views": "platform",
"metrics.reposts": "platform",
"author.verified": "capability",
"author.followers": "capability",
"hashtags": "capability",
"mentions": "capability",
"is_pinned": "capability",
"post_id": "platform",
"parent_id": "platform",
"root_id": "platform"
},
"billing": {
"amount": "0.02",
"currency": "USD",
"balance_after": "9.98"
}
}
}From a response recorded 2026-10-12 18:43 UTC. Names, text and IDs are examples, and so are the links: none is the recorded one (the page shows no one's personal data); the figures are as recorded. Long lists are cut. The request ID, cursor and balance are made up.
Xiaohongshu note search · $0.0200 / a call · Parameters, fields and errors on the endpoint's page →
Will a data API get my account banned?
It does not use your account: a call carries only DataXAPI's key, with no platform account, cookie or proxy of yours.
Can scraper code be switched to the API?
Mostly one to one: search, details, comments and an account's posts are an endpoint each; for the next page pass page.next_cursor back as cursor; the author is always in author and counts in metrics, so another platform needs no new parser.
Is the data current?
Calls by ID, keyword or account are fetched at call time, and every response carries meta.fetched_at, the time the data was read.
What to watch
Collect only what is public. Copying by hand, a plugin and a scraper all sign in with your own platform account, and going fast or taking much gets it limited, at worst banned; a data API uses no account of yours. Being able to read something is not a right to republish it: keep to the law and the author's rights, and take more care where personal information is in it.
Endpoints
Each with its price; every page has the parameters, the fields and a recorded response.
Sources
What this page says of other services is from their own pages and documents, read on 2026-11-20.
- MediaCrawler: Xiaohongshu client
- MediaCrawler issue 757: 461 and 300011
- f2 (open-source project): Douyin utilities
- yutto issue 876: -352 and 412
- weibo-crawler issue 565: what 432 is
- wechat-article-exporter's README