Xiaohongshu (RedNote) API: note text, image OCR, comments and video transcripts

Short answer: Xiaohongshu (RedNote, 小红书) has no public API for reading notes — its open platforms serve shops, mini-programs and advertisers. To get a note into code or an LLM you read its public share page. LinkDigest does that as one call: send a note link (or the whole share text the app copies) and get back the caption, every image described and its text read, the transcript of a video note, the top comments and the like/collect/comment counts, as JSON or Markdown, with no Xiaohongshu account and no cookies.

Disclosure: LinkDigest is the author's own service.

What comes back

Field	What it holds
`title`, `caption`, `tags`, `author`, `posted_at`	The note as posted
`images[]`	Each image described in words, and the text written on it (OCR) — on Xiaohongshu that text is usually the content
`transcript[]`	For a video note: `{t, text}` segments, `t` in seconds
`ocr_text` / `on_screen[]`	Text burned into the video frames, with times
`comments[]`	The top comments the share page shows, with like counts
`stats`	Likes, collects, comments, shares as of the read
`key_points` + `key_point_support`	Short points, each backed by a quote checked word for word against the note
`coverage`	What was read and what was not (images, seconds, comments)

One call

curl -X POST https://linkdigest.dev/api/v1/digest \
  -H "Authorization: Bearer ld_live_..." \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.xiaohongshu.com/discovery/item/...?xsec_token=...", "format": "json"}'

import requests

r = requests.post(
    "https://linkdigest.dev/api/v1/digest",
    headers={"Authorization": "Bearer ld_live_..."},
    json={"url": "http://xhslink.com/a/...", "format": "json"},
).json()
for img in r["images"]:
    print(img["ocr"])

Short links (xhslink.com, xhslink.cn), rednote.com and the pasted share text all work. Video notes may return a job id first; collect it with GET /api/v1/digest/{job_id}. For agents, the same read is the digest_url tool of the MCP server at https://linkdigest.dev/mcp (Claude Code, Cursor, Cherry Studio).

More than one note

Search a topic: POST /api/v1/search (or search_links) finds public notes the web has indexed — not the in-app search — and digest: true reads them.
Compare a batch: up to 20 links in compare_links, or a batch with synthesis: true, return one page: shared themes, hooks, formats, reader concerns, each with its source.
Consumer insight: insight: true adds audience, claims, objections, purchase intent, products and prices, and readers' questions, each with a checked quote.

Limits, said plainly

Use the share link. Xiaohongshu shows a "page not found" screen to a bare note address without its xsec_token as often as not; the link the app's Share button copies carries it.
Some notes are not served at all. Xiaohongshu answers some notes with "当前笔记暂时无法浏览" (this note can't be viewed right now) to every web reader; those are refused and not charged.
Comments: the top comments on the share page, not the full thread.
Private and deleted notes cannot be read.

Price

10 free credits on sign-up, no card. A note with up to six images is 1 credit; links anyone has read before are free, and reads that fail are free. Packs are one-time: 50 credits for $1.50 or ¥9.9 by Alipay. More: Xiaohongshu coverage · why ChatGPT can't open these links · the two share-link domains. Details: pricing · docs · Chinese social media API.

This is the write-up behind LinkDigest, which does the above as a hosted service — with an MCP server for Claude Code and Cursor, a REST API, and 10 free credits to try it.