TikTok data.
Without the scraping.

Here it means one HTTP endpoint that returns public TikTok data as clean JSON: an account’s profile and posts, one video in full, keyword search, the videos under a hashtag, a video’s comments, and the transcript TikTok publishes for a video. You send a handle, a URL, a term or a tag with a Bearer token, and get back typed records with captions, timestamps, durations, engagement counts, authors, music and hashtags. It is not TikTok’s own developer platform — there is no application to file, no review, and no OAuth. Everything returned is what an unauthenticated visitor to tiktok.com can see.

No. You send your FetchLayer key and nothing else. There is no TikTok login, no app registration and no approval step, because this reads the public site rather than a partner API. FetchLayer is not affiliated with, endorsed by, or sponsored by TikTok, and the TikTok name is used only to identify the platform the data comes from.

The same fields on every route that returns videos, so code written against a search result works unchanged on an account’s posts or a hashtag feed: id, url, description (the caption), createdAt, durationSeconds, isPhotoPost, coverUrl, assets, stats with playCount, diggCount, commentCount, shareCount, collectCount and repostCount, author, music, hashtags, mentions, isAd and language. Hashtags and mentions come from TikTok’s own annotations on the post rather than from parsing the caption, so a tag written inside a word does not produce a false positive. A null counter means TikTok did not publish that figure; a real zero comes back as 0.

No. TikTok publishes both the rounded figure shown on the page and the precise one, and this returns the precise pair — a follower count comes back as 1719888 rather than 1.7M. That is the difference between being able to track week-to-week movement and not: a creator can gain thirty thousand followers without 1.7M changing at all.

No, and the API says so rather than pretending otherwise. TikTok offers a public caller neither a date filter nor a sort order on search, so passing one returns a 400 naming the field instead of being accepted and quietly ignored — which would hand you the wrong data with no indication why. Search results come back in TikTok’s own ranking for the term. An account’s own videos can be sorted: latest, popular or oldest.

limit caps the items returned and pages caps how many pages are read to get them, up to 20 in one request. When more exists, hasMore is true and nextCursor carries an opaque value to pass back as cursor. Cursors belong to the route and query that produced them, so do not carry one between routes. Passing -1 on limit or pages means "the maximum this service allows", which is 500 videos, 500 comments and 20 pages.

One credit per page of results read, and every response states it as pagesFetched, so a bill is never a surprise. A single-page request costs 1 credit whatever the page holds — thirty videos or five. Asking for five pages costs five. Parsing a URL with /resolve-url reaches TikTok not at all, reports pagesFetched 0 and is free. TikTok is in Beta while coverage expands, billed at half credit.

Yes, and the honest figure is 68% — 125 of 185 videos measured across 14 languages carried a transcript. TikTok transcribes spoken audio for most videos but not all, and a photo carousel is 0% by nature because there is no audio to transcribe. What makes the route worth more than the number suggests is that the translations are free: a video posted in another language usually carries an English track alongside its original, already timed, for the same one credit. You get the full text plus segments with start and end times in seconds. A video without one is a normal 200 — transcript.available is false, tracks is empty, durationSeconds is still filled in, and a note says why — so do not retry it.

The comments themselves page normally, up to 500 in one request. Two things are worth knowing. The replies array on a comment holds only the replies TikTok shows alongside it — usually none or a handful — while replyCount reports how many exist in total, and a response where the two differ carries a note saying so. And totalCount is the figure TikTok publishes for the video, which can exceed what paging can reach. Both are stated rather than glossed over, so you can tell a complete read from a partial one.

You get an explanation, not an empty array. Every response carries a notes array, and an empty result always comes with a note saying why: the account is private, the account has no public posts, comments are turned off for the video, the hashtag has no public videos, or the list stopped before the requested number was reached. A missing answer and an explained absence are different things, and the response distinguishes them.

Every route is monitored 24/7 and a failure alerts our team, and you are only charged for successful responses. Being honest about the one soft spot: listing an account’s videos is the route TikTok declines most often. In practice it succeeds about 199 times in 200, so the failures are rare but real. A 503 there means the route has genuinely run out of road — wait before retrying rather than retrying immediately.

Yes. All eight routes are MCP tools on the same FetchLayer MCP server, named tiktok_user_profile, tiktok_user_videos, tiktok_video_detail, tiktok_search_videos, tiktok_hashtag_videos, tiktok_comments, tiktok_video_transcript and tiktok_resolve_url. They are distinct from the tiktok_ad_library_* tools, which read the paid ad library rather than organic posts. Tool calls use the same key and are billed exactly like REST.

Account analytics — audience demographics, watch time, traffic sources — are not available from anyone for an account you do not own, because TikTok publishes none of it. There are no historical snapshots either: counts are as they stand when you ask, so tracking movement means storing what you read and asking again.