Xiaohongshu Comments Scraper — RedNote Replies, Authors & IP City 💬

🎉 One flat row per comment and per reply — with the thread linkage on the row

Built for anyone who has tried to open a nested replies[] array in a spreadsheet

---

🔎 What is the Xiaohongshu Comments Scraper — and when should you use it?

The Xiaohongshu Comments Scraper turns Every comment and nested reply under any Xiaohongshu (RedNote) note into clean, structured rows you can filter, export and feed straight into a spreadsheet, database or AI agent. No account, no portal API key, no browser automation to maintain.

Use it when you need: the comment thread under a specific RedNote note, in the original Chinese, with one flat row per comment and per reply. Each row names the thread root, the exact comment being answered and that comment's author. Alongside that sit the commenter's display name, Xiaohongshu ID, public Red ID, avatar, profile link and like count. The IP city RedNote publishes beside a comment comes through where it is published, as do any full-size attached pictures, @-mentions and hashtags. Three booleans decoded from the platform's own tag array say whether the note's author wrote the comment, whether they liked it, and whether it is pinned.

Use something else when: you want the notes themselves rather than the conversation under them. Use Xiaohongshu Scraper - Notes, Users & Search for note details, creator profiles, a creator's full published back catalogue, and keyword search across notes and users. Use Xiaohongshu KOL Analytics for creator-marketplace numbers for a RedNote KOL: audience profile, cost-effectiveness and content tags. This actor reads comment threads on Xiaohongshu only, and only on notes that are publicly visible.

🤖 Use with AI agents

Already connected to the Apify MCP server? Just ask for this Actor by name: sian.agency/xiaohongshu-rednote-comments-scraper

Your agent can pay for its own runs. This Actor is eligible for agentic payments, so an agent can discover it, run it and settle the bill over x402 (USDC on Base) or Skyfire — without an Apify account or API token of its own. Billing is the same either way: per successful row, never for errors.

Otherwise copy this prompt into Claude, ChatGPT, Cursor or any MCP-enabled assistant:

I want the comment and reply threads under specific Xiaohongshu (RedNote) notes using the Apify Actor `sian.agency/xiaohongshu-rednote-comments-scraper`.

Use it when I need: the comment thread under a specific RedNote note, in the original Chinese, with one flat row per comment and per reply. Each row names the thread root, the exact comment being answered and that comment's author. Alongside that sit the commenter's display name, Xiaohongshu ID, public Red ID, avatar, profile link and like count. The IP city RedNote publishes beside a comment comes through where it is published, as do any full-size attached pictures, @-mentions and hashtags. Three booleans decoded from the platform's own tag array say whether the note's author wrote the comment, whether they liked it, and whether it is pinned.

Don't use it when: you want the notes themselves rather than the conversation under them — use xiaohongshu-rednote-scraper or xiaohongshu-kol-analytics instead.

How to call it: put one or more note references in `noteUrls` — a browser URL (`https://www.xiaohongshu.com/explore/...` or `/discovery/item/...`), an `xhslink.com` share link copied from the phone app, or a bare 24-character note ID, mixed freely in one list; set `sortBy` to `latest` for newest-first or `normal` for RedNote's own ranking; leave `includeReplies` on to get every reply as its own row and raise `maxRepliesPerComment` to 0 for whole threads; bound the spend with `maxCommentsPerNote` (top-level comments per note, 0 for all) and `maxResults` (the hard row ceiling for the run). There is no like-count sort: the source accepts one and then ignores it, so sort on the `likeCount` column after export instead.

Start with this input:
{
  "noteUrls": [
    "https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c"
  ],
  "sortBy": "latest",
  "includeReplies": true,
  "maxRepliesPerComment": 20,
  "maxCommentsPerNote": 100,
  "maxResults": 1000
}

Ask me which note links to read, and whether reply threads should be included or only the top-level comments, then run the Actor and summarise the results as a table.

Things you can ask your agent for:

• Pull every comment on these five competitor launch notes and group the complaints by theme.
• Find the comments the creator replied to herself, and show me what she said back.
• Rebuild the longest reply thread on this note in order, so I can read the argument end to end.

Machine-readable API, MCP config and OpenAPI definition for this Actor are published at apify.com/sian.agency/xiaohongshu-rednote-comments-scraper.md.

📋 Overview

Comment threads are where Chinese consumers say what they actually think. A RedNote note is a polished post. The 300 comments underneath it are a focus group nobody moderated, and this Actor exports all of them.

What you get:

• ✅ Flat rows, not nested arrays: every reply is its own row carrying parentCommentId, replyToCommentId and replyToAuthorName. Open the CSV in Sheets and it just works.
• 🧵 Reply threads that go deep: set maxRepliesPerComment to 0 and a 200-reply argument comes back whole, page by page, with no duplicates.
• 🎯 Signals the platform buries: whether the note's author wrote the comment, whether they liked it, whether it is pinned. Three booleans, decoded from a tag array most tools hand back raw or discard.
• 💰 Charged per row returned: a note with no comments costs you nothing. Turn replies off and the reply charge never fires.
• 📍 IP city where RedNote publishes it: the province or city shown beside the comment, filled whenever the commenter's settings allow it.
• 🖼️ Comment pictures at full size: unboxing shots and defect photos come through as image URLs, not thumbnails.

---

✨ Features

• 🔗 Three input formats: browser URL, xhslink share link from the phone app, or a bare 24-character note ID. Mix them in one list.
• 🧵 Nested reply expansion: replies are fetched thread by thread and flattened, so a conversation rebuilds with a single group-by.
• ↩️ Reply-to attribution: inside a long thread, a reply usually answers another reply rather than the root. Both links are on the row.
• 🔀 Two comment orders: newest-first for monitoring, or the ranking RedNote itself shows.
• 📌 Pinned, author-written and author-liked flags as plain booleans.
• 🏷️ @-mentions and hashtags split out of the comment text into their own arrays.
• 👤 Full author block: display name, Xiaohongshu ID, public Red ID, avatar and a direct profile link.
• 📊 Thread totals on every row: how many comments the note has in all, and how many are top-level, so you can tell what a capped run left behind.
• 🛑 Two spend caps: per note and per run, both set before anything is fetched.
• 🚫 No login, cookies or API key anywhere in the process.

---

🎬 Quick Start

Paste one or more note links, press Start, and read the dataset. Everything else has a working default, so a first run needs exactly one field. Results land as JSON, CSV or Excel.

curl -X POST "https://api.apify.com/v2/acts/sian.agency~xiaohongshu-rednote-comments-scraper/runs?token=YOUR_TOKEN" \
-H 'Content-Type: application/json' \
-d '{"noteUrls": ["https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c"]}'

---

🚀 Getting Started (3 Simple Steps)

Step 1: Copy the note link

Open the note in a browser and copy the address bar, or hit Share in the app and copy the xhslink.com link. Both work.

Step 2: Paste it into Note URLs

Add as many notes as you like. Use Bulk edit to paste a whole list, one per line.

Step 3: Press Start

Leave the rest alone for a first run, or set maxRepliesPerComment to 0 if you want every reply of every thread.

That's it. Within a minute or two you'll have:

• Every top-level comment, newest first
• Every reply, linked to the comment it answers
• Authors, avatars, IP cities, likes and any attached pictures

---

📥 Input Configuration

Field	Type	Required	Description
noteUrls	array	No	Note URLs, xhslink share links or 24-character note IDs
sortBy	string	No	latest (newest first) or normal (RedNote's own ranking)
includeReplies	boolean	No	Fetch reply threads. On by default
maxRepliesPerComment	integer	No	Replies per thread. 0 takes every reply
maxCommentsPerNote	integer	No	Top-level comments per note. 0 takes the whole thread
maxResults	integer	No	Hard ceiling on rows for the whole run

Example:

{
  "noteUrls": ["https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c"]
}

Several notes, every reply, capped spend:

{
  "noteUrls": [
    "https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c",
    "http://xhslink.com/a/EXAMPLE",
    "69ec85eb000000001e00cd6c"
  ],
  "sortBy": "latest",
  "includeReplies": true,
  "maxRepliesPerComment": 0,
  "maxCommentsPerNote": 0,
  "maxResults": 5000
}

Top-level comments only, cheapest mode:

{
  "noteUrls": ["6a79fb020000000027023c9c"],
  "includeReplies": false,
  "maxCommentsPerNote": 200
}

---

📤 Output

One row per comment and one row per reply, with 30+ fields on each:

Field	Type	Description
commentId	string	The comment's own ID
content	string	What was written, in the original Chinese
likeCount	integer	Likes on this comment or reply
ipLocation	string	Province or city RedNote shows, where published
postedAt	string	ISO timestamp, UTC
isReply	boolean	True for a reply row
parentCommentId	string	The thread root — group on this
replyToCommentId	string	The exact comment being answered
replyToAuthorName	string	Who is being answered
replyCount	integer	Replies under a top-level comment
authorName	string	Who wrote it
authorRedId	string	Their public Red ID
authorAvatarUrl	string	Profile picture
authorProfileUrl	string	Direct link to their profile
authorIsNoteAuthor	boolean	The creator answering their own audience
likedByNoteAuthor	boolean	The creator liked this comment
isPinned	boolean	Pinned to the top of the thread
pictureUrls	array	Full-size images attached to the comment
mentionedUsers	array	Accounts tagged in the text
hashtags	array	Topic tags used in the text
noteCommentTotal	integer	Comments on the note in total

Example — a reply:

{
  "commentId": "6a956f660000000015010a98",
  "content": "为啥我一搜都是一百多的",
  "likeCount": 0,
  "postedAt": "2026-08-31T12:11:18.000Z",
  "ipLocation": "河南",
  "isReply": true,
  "parentCommentId": "6a9534a200000000020121e3",
  "replyToCommentId": "6a9534a200000000020121e3",
  "replyToAuthorName": "优雅的不纳拿",
  "authorName": "巫师Qiya",
  "authorRedId": "63611470514",
  "authorProfileUrl": "https://www.xiaohongshu.com/user/profile/aba20d57cda73d0896d49573b7ed3",
  "authorIsNoteAuthor": false,
  "likedByNoteAuthor": false,
  "isPinned": false,
  "pictureCount": 0,
  "hashtags": [],
  "noteId": "6a79fb020000000027023c9c",
  "noteUrl": "https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c",
  "noteCommentTotal": 330,
  "status": "success"
}

---

💼 Use Cases & Examples

1. China market entry research

A brand manager wants to know how Chinese consumers actually talk about a product category before committing to a launch.

Input: the twenty most-engaged notes in the category Output: several thousand comments and replies, with IP city where published Use: an unmoderated focus group, in the customer's own words, for a fraction of what an agency panel costs

2. Brand and campaign monitoring

An agency runs a KOL campaign and needs to know what happened underneath each sponsored note.

Input: the campaign's note links, on a daily schedule Output: every comment and reply, with pinned and author-liked flags Use: the flags reveal which reactions the creator chose to amplify. That is a different signal from raw like counts, and one no screenshot gives you

3. Training data for Chinese-language models

An ML team needs conversational Chinese with real turn-taking, not scraped articles.

Input: a list of high-comment notes across topics Output: rows where every reply names the comment it answers and that comment's author Use: dialogue trees rebuild without guessing who was talking to whom

4. Competitor product research

A product team wants the complaints under a competitor's launch posts.

Input: the competitor's note links Output: comment text plus full-size attached pictures Use: defect photos, side-by-side comparisons and unboxing shots come through as image URLs instead of being lost with the rest of the payload

5. Community management and moderation review

A creator or brand needs a record of everything said under their own notes.

Input: their own note links Output: the complete thread, replies included Use: audit what was said, spot spam patterns in the mention and hashtag columns, and keep an export the RedNote interface will not give you

6. KOL vetting before a partnership

A media buyer wants to know whether a creator's audience is real and engaged.

Input: several notes from the creator under review Output: comment authors, their Red IDs, IP cities and like distribution Use: a thread of one-word comments from accounts with no history reads very differently from a genuine discussion

---

🔗 Integration Examples

JavaScript/Node.js

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: 'YOUR_TOKEN' });

const run = await client.actor('sian.agency/xiaohongshu-rednote-comments-scraper').call({
  noteUrls: ['https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c'],
  includeReplies: true,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();

// Rebuild one thread
const threads = {};
for (const row of items) {
  const key = row.parentCommentId || row.commentId;
  (threads[key] ??= []).push(row);
}

Python

from apify_client import ApifyClient
client = ApifyClient('YOUR_TOKEN')

run = client.actor('sian.agency/xiaohongshu-rednote-comments-scraper').call(
    run_input={
        'noteUrls': ['https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c'],
        'includeReplies': True,
    }
)

for item in client.dataset(run['defaultDatasetId']).iterate_items():
    prefix = '    ↳ ' if item.get('isReply') else ''
    print(prefix, item['authorName'], item['content'])

cURL

curl -X POST 'https://api.apify.com/v2/acts/sian.agency~xiaohongshu-rednote-comments-scraper/runs?token=YOUR_TOKEN' \
-H 'Content-Type: application/json' \
-d '{"noteUrls": ["https://www.xiaohongshu.com/explore/6a79fb020000000027023c9c"]}'

Automation Workflows (N8N / Zapier / Make)

1. Trigger: a schedule, or a webhook when a new campaign note goes live
2. HTTP Request: call the Actor with the note links
3. Process: filter on isReply, or group on parentCommentId
4. Action: push to a sentiment model, a warehouse, or a Slack digest

---

📊 Performance & Pricing

FREE Tier (Try It Now)

• 10 rows per run — every field, same quality, no credit card
• Enough to see the thread linkage working on a real note before you commit

PAID Tier (Production Ready)

• Unlimited rows per run, across as many notes as you like
• Charged per row returned: a note with no comments costs nothing, and error rows are never billed
• Replies are billed separately from comments, so switching them off genuinely lowers the bill

💰 Priced under the category leader on every paid tier — $0.0058 per comment against their $0.00599, and $0.0049 against their $0.00499 floor. The row you get back also needs less work before it is usable.

🔗 View current pricing

---

❓ Frequently Asked Questions

Q: Are replies nested inside the comment, or separate rows? A: Separate rows. Each one carries isReply: true, a parentCommentId pointing at the thread root and a replyToCommentId pointing at the comment it actually answers. Group on parentCommentId to rebuild a thread. Nested arrays are the common approach and they turn into replies/0/content, replies/1/content columns the moment you export to CSV.

Q: Does every comment have an IP city? A: No. RedNote shows it only where the commenter's own settings publish it. Checked on two live notes the same afternoon, one returned an IP city on all 10 comments of a page, the other on 1 of 25. The column is always there and is filled whenever the data is.

Q: Can I sort comments by like count? A: Not honestly, so it is not offered. The source accepts a like-count sort and then returns default order regardless. The response even reports which ranking it fell back to. Rather than ship a control that quietly does nothing, you get newest-first and RedNote's own default. Sort on the likeCount column after export.

Q: How many comments can I pull from one note? A: All of them. Comment pages are walked until the source says there are no more, and reply threads paginate the same way. The caps exist so you can bound the spend; set them to 0 and raise maxResults to take everything.

Q: Which link formats work? A: A browser URL (/explore/... or /discovery/item/...), an xhslink.com share link from the app, or a bare 24-character note ID. A creator profile link will be rejected before anything is charged, because it points at a person and not a note.

Q: Do I need a login, cookies or an API key? A: No. Paste links and press Start.

Q: Will I get sentiment scores or English translations? A: No. You get the comment text exactly as written, in Chinese. Any tool offering a sentiment column here is generating it, not collecting it. Run the content column through your own model and you keep control of how it was scored.

Q: What happens if a note has been deleted? A: It comes back as an error row explaining why, and it is not charged. The rest of your notes carry on.

---

🐛 Troubleshooting

A note returned zero rows

• Open the link in a browser. If the note is gone or has been made private, there is nothing to collect and you were not charged.
• If you can see comments in the browser but not here, send us the link on Telegram.

"That does not look like a Xiaohongshu note reference"

• The link points at a profile, a search page or a topic page rather than one note.
• Copy the address while the note itself is open.

A share link failed

• Share links expire. Open it once in a browser, then copy the www.xiaohongshu.com address it lands on.

Fewer replies than the note shows

• maxRepliesPerComment defaults to 20. Set it to 0 for the whole thread.
• Check maxResults too — it stops the run across all notes, replies included.

The run stopped early on a free plan

• The FREE tier returns 10 rows per run. Add credits or a payment method to lift it.

---

⚠️ Trademark Disclaimer

Xiaohongshu, RedNote and 小红书 are trademarks of Xingin Information Technology Co., Ltd. This tool is not affiliated with, endorsed by or sponsored by Xiaohongshu, and collects only publicly visible comments.

---

⚖️ Is it legal to scrape data?

Our actors are ethical and do not extract any private user data, such as email addresses, gender, or location. They only extract what the user has chosen to share publicly. We therefore believe that our actors, when used for ethical purposes by Apify users, are safe.

However, you should be aware that your results could contain personal data. Personal data is protected by the GDPR in the European Union and by other regulations around the world. You should not scrape personal data unless you have a legitimate reason to do so. If you're unsure whether your reason is legitimate, consult your lawyers.

You can also read Apify's blog post on the legality of web scraping.

---

🤝 Support

Join our active support community

• For issues or questions, open an issue on the Actor's Issues tab
• Check SIÁN Agency Store for more automation tools
• 📧 apify@sian-agency.online

---

Built by SIÁN Agency | More Tools