WeChat Universal Search Scraper

EN — Find WeChat content across 公众号 / 视频号 / 文章 / 账号 in 30 seconds. Drop one keyword, get up to 50 ranked results (5 pages × ~10) with engagement metrics (likes, comments, duration), publisher profile, shareable URLs, and cover images. No login, no captcha, no headless browser.

ZH — 微信「搜一搜」综合搜索 — 公众号·视频号·文章·账号 一键抓取。 一个关键词,最多 50 条结果(5 页 × ~10),含点赞数/评论/时长、发布者资料、可分享链接、封面图。无登录、无验证码、无 headless 浏览器。

Built for cross-border e-commerce sourcing, influencer discovery, trend monitoring, and KOL prospecting on WeChat — China's #1 super-app with 1.3 billion monthly users. Output exports as JSON / CSV / Excel — pipe directly into your CRM, BI tool, or another Apify Actor.

5 free runs / day — no API key required. Open on Apify Console, drop a keyword, get results in 30 seconds.

What you can search

businessType	What you get	EN	中文
all (default)	All verticals mixed	All	综合
account	公众号 / 服务号 / 视频号 accounts	Accounts	公众号账号
article	公众号 articles	Articles	公众号文章
video	视频号 Channels videos	Videos	视频号视频
sticker	表情 stickers	Stickers	表情包
underline / encyclopedia / live_stream / comment / listen / news / photos / book / moments / image / mini_game / weixin_index / ai_search	(TikHub documents these verticals but they currently return empty due to server-side geographic / account gating)	—	实测空

All inputs also support:

• keyword (1-100 chars, e.g. 美食, 人民日报, 跨境电商)
• sort: default (relevance) / latest / hot
• publishTime: all / day (24 h) / week (7 d) / half_year (6 mo)
• maxPages: 1-5 pages, default 3 (cost control)
• cursor: opaque token from a previous run's SUMMARY.nextCursor for resumable pagination
• raw: include the unprocessed upstream data envelope on each pushed item (off by default)

Input

{
  "keyword": "美食",
  "businessType": "video",
  "sort": "default",
  "publishTime": "all",
  "maxPages": 3,
  "cursor": "",
  "raw": "false"
}

Output

Each pushed item is a normalized row keyed by businessType + docId (string — base64 may exceed JS Number safe range):

{
  "keyword": "美食",
  "page": 1,
  "rank": 1,
  "businessType": "video",
  "businessTypeRaw": 7,
  "docId": "finderobjv0iGrW2Ju7cmIl5Q2m0E",
  "exportId": "export/UzFfAEDovIKb9q9zJ4...",
  "hashDocId": "17711004180000001",
  "feedNonceId": "1717...",
  "title": "家常菜快手菜...",
  "titleRaw": "...<em class=\"highlight\">美食</em>...",
  "coverUrl": "https://wx.qlogo.cn/...",
  "duration": "03:42",
  "pubTime": 1757365200,
  "dateLabel": "2小时前",
  "likeNum": "10万+",
  "source": {
    "name": "刘小妹正餐",
    "iconUrl": "https://wx.qlogo.cn/...",
    "verified": true
  },
  "collectedAt": "2026-09-09T00:00:00Z"
}

Article-shaped items add docUrl (公众号原文链接) and description; account-shaped items add description (account intro) and use coverUrl = iconUrl. The businessTypeRaw integer field echoes the upstream business_type code so you can route / segment downstream (7 = video, 2 = article, 33554499 = account — the actual upstream magic number, not 14).

The run also writes a SUMMARY to the Key-Value Store with pushed / pagesFetched / nextCursor / businessTypeRaw / hasMoreInSource / tier / runsToday / dailyLimit / startedAt / finishedAt — use nextCursor on the next run for resumable pagination.

Use cases

• Cross-border e-commerce sourcing / 跨境电商内容监控: find 公众号 + 视频号 KOLs posting about your category, capture exportId (video) or docUrl (article) to download the media, push to your CRM.
• Influencer discovery (视频号) / KOL 调研: filter by businessType=video + publishTime=week to surface emerging Channels creators before they saturate.
• Trend monitoring / 舆情追踪: run daily with the same keyword, diff docId sets between runs to spot new entrants.
• Brand monitoring / 品牌账号监测: track 公众号 articles mentioning your brand or product category, paginate via nextCursor.
• Account prospecting / 公众号账号挖掘: businessType=account returns account metadata + icon, ideal for building a target publisher list.

Pagination

The Actor auto-paginates via the upstream cursor + continue_flag envelope. Set maxPages (1-5) to cap requests per run (5 pages ≈ 50 items). To resume a previous run, copy SUMMARY.nextCursor from the previous run into cursor on the next run.

maxPages	Items per run (avg)	Cost control / 成本控制
1	~10	smoke test / 冒烟测试
3 (default)	~30	sweet spot / 推荐值
5	~50	hard cap (cost & quota safety)

Quick start

Open the Actor on Apify Console, fill in keyword (e.g. 美食), click Start. Results land in the Dataset tab in ~30 seconds.

Want to run from the CLI?

$apify call wechat-universal-search-scraper -i '{"keyword": "美食", "businessType": "video"}'

The upstream API key is wired into the Actor — no Secrets tab setup, no env vars to set. Apify free users get 5 runs / day.

Sister Actor — Channels video search

For 视频号 Channels-only video search with duration filter (short / medium / long) and sort (latest / hottest), see the sister Actor:

→ WeChat Channels Video Search Scraper — same keyword, dedicated video endpoint with time / length filters.

Tests

$npm test

Runs 23 unit assertions (input validation, request body shape, normalization across video / article / account shapes) plus 7 integration scenarios (mock fetch drives the full run() flow against captured upstream fixtures).

✅ unit.js: 23 assertions passed.
✅ integration.js: 7 scenarios passed.

Notes on upstream behavior

• business_type is an integer code per upstream / business_type 是上游整数代码: 7 = video (视频号), 2 = article (公众号文章), 33554499 = account (the actual upstream magic number — TikHub's docs mention 14 for account but the live API returns 33554499). The Actor accepts the upstream's code verbatim and exposes it as businessTypeRaw; the matched businessType string comes from your input.
• 13 verticals return empty pages / 13 个分类返空: TikHub documents 18 business_type values but only 5 (all / account / article / video / sticker) currently return real results; the other 13 are server-side gated by geography / account. The Actor exposes all 18 per AGENTS-apify-archive.md §40 (open enum = no surprises).
• 30-second hard timeout / 30 秒硬超时: TikHub documents that the upstream takes up to 30 s. The Actor enforces this with AbortController + exponential backoff (3 retries: 500 ms / 1 s / 2 s).
• Large integer IDs preserved as string / 大整数 ID 保 string: docID is base64, hashDocID is a 19-digit number, likeNum includes Chinese units like 10万+. All kept as strings to avoid JS Number precision loss.

Changelog

See ./CHANGELOG.md.

License

Apache-2.0