The social media reading layer for AI agents
From search to extraction, one closed loop. Search tools stop at links, extraction tools wait for a pasted link; LinkDigest does both: find posts on Xiaohongshu, Douyin, TikTok, YouTube and X by topic, then read each one into structured, verifiable data your AI agents can use.
$0.02 an image post, $0.02 per minute of video, 10 free on sign-up.
New Pain-Point Radar: what buyers complain about and search for, counted across the top posts. See the radar · For business: the competitor report
One free preview, no account: a short post reads whole, a video its first minutes, and signing in (10 free credits) reads the rest. The examples below run instantly because someone has already digested them — cached links are always free.
Try an example below, or paste your own link.
xiaohongshu · a real read
16 images, each described and read
douyin · a real read
transcript + on-screen text from a 2:22 video
youtube · a real read
426 transcript segments, 14 key points
tiktok · a real read
67 native-caption segments
Each one reads posts into the same verifiable data: transcripts, the text in images and video, top comments, key points with their sources.
Xiaohongshu, Douyin, Bilibili, WeChat articles, YouTube, TikTok, X, Weibo and podcasts: the transcript with timecodes, on-screen and in-image text, the top comments a share page serves, key points with checked quotes. A link that will not read costs nothing; a link someone already read is free.
Median whole read, last 30 days of real requests
| Xiaohongshu | 36 s |
| Douyin | 72 s |
| TikTok | fewer than five reads |
| YouTube | fewer than five reads |
| X | fewer than five reads |
Xiaohongshu · Douyin · Bilibili · Your own tab, with the extension
It reads what buyers type into search boxes, then the top posts and their comments on TikTok and YouTube (or RedNote and Douyin), groups what people complain about, want and hesitate over, and counts in code how many posts each point appears in, with a quote behind every line. Searches no post answers become opportunities; a weekly re-run says what changed.
Live run · robot vacuum · 2026-11-07 · 11 posts
How it works · Run one · For business: the competitor report
Ask it the way you would ask a colleague: 「我想在槟城开咖啡馆,顾客怎么评价现在的同行?」 It searches 小红书, 抖音, YouTube and X for real posts, reads them, and comes back with what customers praise, what they complain about and which need nobody serves — every point tied to its post — and a closing 「这周就做」: one to three things to do this week.
Ready-made jobs carry a specialist’s method, and more are added as they come up; or write your own in Skills. In 中文 or English. 2 credits an answer, plus the searches and reads it makes.
Name a niche: each morning the hottest new posts, why they worked, and a draft written from them in your own voice, every figure taken from the notes. Or run a batch of links into one table and one comparison: titles, captions, key points, transcripts, the breakdown (hook, beats, template) and consumer insight.
Batches of 10 links on the free plan, 50 on Dev and 200 on Studio; a parallel-read hold keeps a burst from spending more than your balance.
One Xiaohongshu note read with insight and breakdown, a public video digest for the transcript, and one batch synthesis — replayed from production output, every quote carrying the engine’s own check.
title我确信,有什么东西已经来到了下一个阶段。
authorExplosions(boom · xiaohongshu · 2026-04-26
tags#世界的主人 #韩国电影 #女性
images · 2 described and read
image 1
The first image is a social media post or note with a light purple gradient background. It features large Chinese text that reads 'After watching 'The Master of the World', I dragged my friend and we talked on the phone for 1 hour and 50 minutes'. The date 'APR.27' is visible in the top right corner, and the word '主人' (master/owner) in the title is highlighted with a green underline.
APR.27 看完《世界的主人》 我拉着朋友 打了1小时50分钟的 电话
image 2
The second image displays a block of text, which appears to be a detailed movie review or reflection. The text discusses the director's observation of life, the reconstruction of daily life after trauma, the contrast between main and supporting characters, and praises the film's structure and self-reflexive nature. It mentions specific elements like a car wash scene and a final monologue as being the best of the year.
举轻若重,重建日常。 导演对于观察生活已经不仅仅是可以将其细致入微 地合适地呈现,更是可以从中转化出正视与面对的 力量。审判之后呢?创伤之后呢?所有人只好缝缝 补补、小心翼翼地重建与维护日常。 配角与主角之间的对照也非常耐人寻味,有家庭结 构上的,也有事件进程上的,都非常有血有肉。 很久没有一部在放到第28分钟的时候就让我感到 “太神了”的电影,此时还只是结构上的,后面缓缓 展开的主题更让人反思和感动。 特别喜欢这种不把观众当傻子的电影,片中的人物 没有一五一十地向观众讲述故事全过程的义务,他 们有自己的行事逻辑。 影片结束之后回发觉导演选择这种拍摄方式也是表 明了立场,她坚定地站在珠仁这边,以这样的电影 拍法来帮助其完成对日常的重建。从这一层中感受 到了一些结构上的自反性,一些自指的味道。洗车 与最后的念白是年度最佳。
key_points · 9, each with its quote checked
caption
在豆瓣上写的影评放在了p2。 在上周六快凌晨两点看完《世界的主人》,一直到早上五点多我都没有办法平复我的心情。看完之后我就拉着在香港早早看过的朋友打了一通一小时五十分钟的电话,疯狂输出我想到的理解和评价。我甚至录了下来想日后做成播客或整理成公众号文章。 怎么会有这么好的电影,怎么会有这么好的电影,全方位的好,从电影技法到主题内容都是浑然天成的,举轻若重,平地惊雷,尹佳恩太厉害了。在此地,任何相关议题什么都拍不了,而《世界的主人》已经完成了对于《熔炉》等其本土相关电影的解构,已经是完完全全的next level了。 最近一直在领悟作品当中“日常”这个概念,日常是有很大的力量的,一旦被摧毁,重建日常是非常困难但也是非常必要的。还有太多想说的了,像父亲的缺席其实是一种在场(两组家庭的父亲都是缺失的),各个角色之间的对照关系,说不完。洗车那一段的设计和表演以及最后的念白可以直接预定年度最佳。 我十分夸张地跟朋友说《非穷尽列举》和《世界的主人》给我带来的震撼不亚于上高中的时候第一次了解metoo,前两者是令人感叹本世纪的剧本已经到这么下一层次的阶段了,后者则是让当时的我清晰地认识到前沿的运动已经在把大的叙事目光重新放回到个人身上。它们都让我产生了“有什么东西已经来到了下一个阶段”的感觉。 #世界的主人[话题]# #韩国电影[话题]# #女性[话题]# Tags: 世界的主人, 韩国电影, 女性
The agent calls digest_url by itself when it meets a social link. Links come from search_links (with digest: true the call that found them reads them too), from the extension’s button on a page you can see, or from a paste.
MCP · Claude Code
claude mcp add --transport http linkdigest \ https://linkdigest.dev/mcp \ --header "Authorization: Bearer ld_live_..."
REST
curl -X POST https://linkdigest.dev/api/v1/digest \
-H "Authorization: Bearer ld_live_..." \
-d '{"url": "https://v.douyin.com/..."}'Fetching other people’s platforms breaks, so this table is kept honest rather than aspirational. When a platform blocks us, LinkDigest says so instead of returning a confident description of an error page.
| Platform | State | Median read | Notes |
|---|---|---|---|
| xiaohongshu 小红书 | live | 36 s | Image notes (up to 60 images) and video notes. No login needed. |
| douyin 抖音 | live | 72 s | Pasted links read again since 2026-10-21 through Douyin's own embed player. Video and 图文; long videos transcript-first. Anything it will not serve: upload the video saved in the app (保存本地) at /app/upload, or read it from your own tab with the extension. |
| weibo 微博 | live | — | Post links read through Weibo's own visitor page — no login. Text, every photo, video, reposts. A post only signed-in followers can see (关注可见 / 好友圈) says so instead of guessing and is not charged; if it carries a video, the extension reads it from your own signed-in tab, the way a Douyin video is read. |
| hongguo 红果短剧 | live | — | A drama's page (hongguoduanju.com/detail?series_id=…) and its series player page (/player/<id>, the address an episode opens at) read from the page's own data: synopsis, tags, episodes, rating, likes / saves / heat, its rank label, filing number and the top written reviews. The episode itself is not read — the player ships its tracks CENC-encrypted, so reading one would mean defeating its content protection. For an episode: play it in your own browser (the web player shows the first few free) and read it with the extension. What is winning right now, counted over 48 ranked dramas: https://linkdigest.dev/hongguo-short-drama-trends |
| youtube | live | — | Read by Gemini watching the video, since YouTube blocks datacenter IPs. |
| tiktok | live | — | Short links resolve. Speech is read from the video itself, not the sound track. |
| x | live | — | Video posts in full. Image posts: every photo described. Long-form articles: the lead image. |
| wechat 公众号 | live | — | Public-account articles (mp.weixin.qq.com). No login needed. |
| web pages | live | — | Article text and metadata. |
| bilibili | live | — | Short links (b23.tv) and full video links, read since 2026-10-24 through Bilibili's own embed player — the public page it publishes for other sites. No login needed. A video that is private, deleted or still under review says so and is not charged. |
| live | — | Reels and video posts are read like any video; photo posts are read from Instagram's own embed page (the one it publishes for other sites), with no login. A carousel gives its first slide; private posts cannot be read. | |
| podcast 播客 (小宇宙, Apple Podcasts) | live | — | 小宇宙 episode links and Apple Podcasts episode links (…?i=…), read from the public episode page and feed with no login, as text: 1 credit per started 2 minutes, no base credit — a 45-minute episode is 23 credits. Paid episodes cannot be read and are not charged. |
| spotify | blocked | — | Not readable: without a login Spotify's player serves a 30-second preview, and we never use anyone's login. Many shows are also on 小宇宙 or Apple Podcasts — paste that link. A Spotify link is answered with this and not charged. |
| ximalaya 喜马拉雅 | soon | — | Not yet: the public extractor we would read it with is broken upstream. Many of its shows are also on 小宇宙 or Apple Podcasts. |
Digests are cached and shared, so a link someone already read comes back instantly and never counts against you.
$0
10 credits when you sign up — once, they do not renew. Every platform, full output, no card. Then 50 credits for ¥9.9 / $1.50, or 250 for ¥36 / $5 — once, Alipay or card, no subscription.
$9/month
500 credits a month — about 500 minutes of video, or that many posts.
Every plan includes the REST API and the MCP server — the free plan too. Finding posts costs a credit or two a search (search_links; the extension in your own browser is free); insight, breakdown and a batch synthesis are priced per link on the pricing page.
Share URLs are tokenised and the content is video or images rather than text. Fetching one returns an app-download shell or a login wall, not the post. There is nothing in the HTML to read.
Add the LinkDigest MCP server, then paste the link. The agent calls digest_url on its own and gets the note's images, on-screen text and caption back as Markdown. Share links with xsec_token work exactly as you copied them.
Yes. Where a platform already publishes captions those are used — they are free and more accurate. Otherwise speech is transcribed with Qwen3-ASR, which returns punctuated Chinese. The transcript comes back as sentences; each sentence's start time is estimated from its place in the audio (marked approx), and on-screen text carries its own approximate times.
No. Media is processed in a temporary directory deleted before the request returns. Only the resulting text digest is stored, so a link someone already digested costs nothing to serve again.
Yes. search_links (and POST /api/v1/search) finds Xiaohongshu, Douyin, YouTube, X and web posts on a topic — 2 credits a Douyin or Xiaohongshu search that finds posts, 1 elsewhere, 0 when nothing is found — and with digest: true reads and compares them in one call. For Xiaohongshu and Douyin that is the posts the web has indexed, each read from its public page or Douyin's own player. For your own feed, open the page in your own browser, press the extension's “Digest the notes on this page” button, and the agent picks the batch up with list_batches. LinkDigest never holds your session.
Through a real search API: a public web search scoped to xiaohongshu.com returns the notes the web has indexed, and each is read from its public page — the one a logged-out phone gets. Xiaohongshu's own search sits behind a login, and the only ways to run that from a server are to hold your session or forge the signatures its app sends — we do neither, for your account's sake and ours. So the web's index is the subset you get from here; for your own feed, open it in your own browser, press the extension's “Digest the notes on this page” button, and the batch is read here. No login, no cookies, no forged signatures, ever.
platform, author, title, posted_at, caption, transcript with timecodes, on-screen text with times, images with descriptions, key_points, engagement counts and hashtags, the video's chapters, category and soundtrack as the platform gives them, the links the post points to (from its text, its QR codes and WeChat's 阅读原文 — listed, never opened), the top comments a Xiaohongshu share page serves with the note, and raw_markdown — plus a breakdown (hook, timed beats, reusable template) or consumer insight (audience, claims, objections, purchase intent, products and prices, each with a checked quote) when you ask for one. Markdown for reading, JSON for pipelines.
No company behind this, so here is the evidence instead of a logo wall.
502 reads by users, 90% came back whole; a whole read took 56 s at the median. Failures on our side count against us; our own accounts are left out; per-platform figures are recomputed hourly.
Published as dev.linkdigest/linkdigest. The registry only grants a domain namespace to whoever controls the domain's DNS, and the proof is public.
The SDK, CLI and skills are MIT on GitHub. Releases are built by GitHub Actions and carry a build attestation you can verify with `gh attestation verify`.
A call is a link, nothing else. No cookies, no passwords, no phone number; the reading happens on our servers, so nothing can get your account flagged.
Media is read and discarded. We never store, serve or redistribute the video or images behind a link; comments and private posts are not read.
Cards and Alipay are handled by Stripe. Card numbers never touch our servers, and the 250-credit pack is a one-off, not a subscription.
Run by one developer in Malaysia. Mail support@linkdigest.dev and a person replies, usually the same day. Domain ownership proof is published too.
Checked independentlymcptry 91/100Official MCP Registry · verified publisherListed on Glama