Xiaohongshu (RedNote) Scraper — Notes, Creators & Comments

Export Xiaohongshu / RedNote data as JSON, CSV or Excel. Search by keyword, pull full note details, profile creators, and collect comments — without an API key, a Chinese business entity, or an agency retainer.

Xiaohongshu has 300M+ monthly active users and serves 600M+ searches a day. It is where Chinese consumers decide what to buy. It is also almost completely closed to outsiders: there is no public API, the web endpoints are cryptographically signed, and full platform access requires a mainland Chinese company. This Actor gives you the data layer without any of that.

---

Quick start (3 minutes)

1. Grab a cookie — open https://www.xiaohongshu.com/ in your browser, log in, press F12 → Console tab, type document.cookie, copy the whole string.
2. Click "Try for free" at the top of this page (or paste the cookie into the cookie field of a run).
3. Hit "Start" — results appear in the Dataset tab within ~30 seconds.

Minimal input that works:

{
  "searchKeywords": ["护肤"],
  "cookie": "a1=...; web_session=...;",
  "maxItems": 20
}

---

What you can do with it

You are	You use it to
A brand entering China	Measure share of voice, find which product claims resonate, track competitors' campaigns
An agency	Build KOC/KOL shortlists from real engagement numbers instead of media kits
A dropshipper / sourcing operator	See what is trending in China 3–6 months before it reaches Western marketplaces
A market researcher	Pull thousands of first-person consumer reviews on any category
A trend / AI team	Feed a genuinely hard-to-obtain Chinese-language dataset into your models

---

Three ways to use it

1 · Search by keyword

The most common use — scrape notes matching a query.

{
  "searchKeywords": ["秋冬护肤", "敏感肌"],
  "cookie": "...",
  "maxItems": 200,
  "sort": "most_liked",
  "noteType": "all",
  "fetchNoteDetails": true,
  "maxCommentsPerNote": 20
}

2 · Deep-scrape specific notes

Already have a shortlist? Paste note URLs — the Actor pulls full body text, tags, all images, IP location, publish time.

{
  "noteUrls": [
    "https://www.xiaohongshu.com/explore/65f2a1b3000000001203c4d5?xsec_token=ABC123"
  ],
  "cookie": "...",
  "maxCommentsPerNote": 50
}

Keep the xsec_token query parameter when copying note URLs — Xiaohongshu refuses the request without it.

3 · Build a KOL shortlist

Pull creator profiles with their recent notes.

{
  "creatorUrls": [
    "https://www.xiaohongshu.com/user/profile/5ff0e6f0000000000101d4e2"
  ],
  "cookie": "...",
  "maxNotesPerCreator": 50
}

---

Input reference

Field	Default	Notes
searchKeywords	[]	Chinese works ×10 better than English — use 护肤, not skincare
noteUrls	[]	Must include xsec_token= query param
creatorUrls	[]	Format: https://www.xiaohongshu.com/user/profile/<userId>
cookie	""	Required in practice. See Cookie walkthrough
maxItems	100	Hard cap. You only pay for records actually delivered
sort	general	general / latest / most_liked / most_commented
noteType	all	all / video / image
fetchNoteDetails	false	Open each note for full body + tags + IP + publish time
maxCommentsPerNote	0	Set > 0 to also pull comments (charged separately)
maxNotesPerCreator	20	Recent notes to collect per creator profile
proxyConfiguration	{useApifyProxy: true, apifyProxyGroups: ["RESIDENTIAL"]}	CN residential strongly recommended

---

Cookie walkthrough

Xiaohongshu heavily restricts anonymous requests. A real browser cookie unlocks full search.

30-second extraction:

1. Open https://www.xiaohongshu.com/ in Chrome / Edge / Firefox
2. Log in (QR code via Xiaohongshu mobile app)
3. Press F12 → Developer Tools opens
4. Click the "Console" tab
5. Paste this and press Enter:
      copy(document.cookie)
6. Your clipboard now holds the cookie string
7. Paste into the "cookie" field of the Actor input

The cookie string looks like:

abRequestId=xxx; a1=xxx; webId=xxx; web_session=xxx; xsecappid=xhs-pc-web; ...

Critical cookies that must be present: a1 and web_session. If either is missing, log in again and re-copy.

Cookies expire. Expect to refresh every 3–7 days under heavy use. The Actor logs a warning when the cookie looks stale.

---

CN proxy — strongly recommended

Xiaohongshu rate-limits aggressively by IP. Without a mainland-China or at least residential proxy, the Actor will often return nothing or get captcha-walled.

Easiest — use Apify's built-in proxy:

"proxyConfiguration": {
  "useApifyProxy": true,
  "apifyProxyGroups": ["RESIDENTIAL"],
  "apifyProxyCountry": "CN"
}

Bring your own — point to any CN residential pool:

"proxyConfiguration": {
  "proxyUrls": ["http://user:pass@cn-proxy.example.com:8080"]
}

---

Output

Each record is a flat JSON object. Chinese counts like 1.2w / 3.4万 / 2亿 are normalised to integers.

Note (from search)

{
  "source": "search",
  "searchKeyword": "护肤",
  "noteId": "65f2a1b3000000001203c4d5",
  "type": "normal",
  "title": "秋冬护肤 10 件小事",
  "description": "...",
  "noteUrl": "https://www.xiaohongshu.com/explore/65f2a1b3000000001203c4d5",
  "xsecToken": "ABC123...",
  "coverUrl": "https://sns-webpic-qc.xhscdn.com/...",
  "likedCount": 12000,
  "collectedCount": 800,
  "commentCount": 36,
  "shareCount": 55,
  "authorId": "5ff0e6f0000000000101d4e2",
  "authorNickname": "护肤小助手",
  "authorAvatar": "...",
  "authorUrl": "https://www.xiaohongshu.com/user/profile/..."
}

Note detail (with fetchNoteDetails: true)

Adds publishedAt, lastUpdatedAt, ipLocation, imageUrls[], videoUrl, tags[], atUsers[], description (full body).

Creator profile

Adds redId (Xiaohongshu ID), bio, gender, ipLocation, followsCount, fansCount, likesAndCollectsCount, tags[].

Comment

{
  "source": "comment",
  "noteId": "65f2a1b3000000001203c4d5",
  "commentId": "1234...",
  "content": "太好用了",
  "likeCount": 42,
  "subCommentCount": 3,
  "createdAt": 1727500000,
  "ipLocation": "北京",
  "authorId": "...",
  "authorNickname": "..."
}

---

Pricing

Pay-per-event — you are only charged for records actually delivered.

Event	Price (USD)	What triggers it
apify-actor-start	$0.005	Once per run (per GB RAM, min 1)
note-scraped	$0.0015	Each note delivered to the dataset
creator-scraped	$0.004	Each creator profile
comment-scraped	$0.0004	Each comment

Cost estimator

Job	Setting	Approx cost
Quick keyword scan	100 notes, no details	$0.16
Deep category sweep	1 000 notes + details + 20 comments each	$9.50
10-KOL shortlist	10 creators × 50 notes each	$0.80
Daily brand-mention monitor	500 notes/day for 30 days	$22.50 / month

Platform margin: Apify takes 20%, you keep 80%. The prices above are what you pay.

---

Common errors

Error / symptom	Cause	Fix
No notes captured for 'X'	Cookie expired / missing	Re-copy cookie as above
Only 1-2 notes returned when more exist	Rate-limited by IP	Enable CN residential proxy
BrowserType.launch timeout	Memory too low	Raise minMemoryMbytes to 4096
Captcha page screenshots in logs	IP blocked	Switch proxy group, wait 10 min
Note URL "refused"	Missing xsec_token= in URL	Copy URL again from Xiaohongshu web, not mobile share

---

Why this exists (and why we don't sign anything)

Xiaohongshu signs every API call with x-s / x-t headers derived from obfuscated browser JS that changes without notice. Re-implementing the signature is a losing maintenance race — Xiaohongshu updates it, your Actor breaks, your users churn.

Instead the Actor opens a real Chromium page in the Apify runtime, lets the page's own JS make the real signed calls, and reads JSON off the wire. Xiaohongshu can rotate signatures all they want — the browser catches up for us.

---

Known limits — be honest with yourself

• Cookies expire. Users are expected to refresh their own cookie. The Actor does not attempt to re-authenticate.
• Search depth is finite. Xiaohongshu caps keyword search around a few hundred results per query — no amount of scrolling pulls more.
• English keywords mostly return nothing. The platform is Chinese-first.
• Comments are heavily paginated. maxCommentsPerNote above a few hundred will slow substantially.
• This is a scraper, not a CRM. Data is read-only, best-effort. Fields may be null or "" when the source didn't provide them.

---

Changelog

• 0.1.8 · Loosen page.goto wait condition + raise timeout to 90s for cross-border runs
• 0.1.7 · Pin pydantic<2.12 to avoid upstream crawlee incompatibility
• 0.1.4 · Add output schema for Store publishing
• 0.1.1 · Initial public release