TikTok Scraper API for Creator and Trend Research

A TikTok scraper API is most useful when it is treated as a retrieval workflow, not a generic content dump. Creator research, trend monitoring, and social signal analysis all need different sequences of calls. A brand analyst may start with keyword search. A creator tool may start with a profile. A music or hashtag monitor may start with a sound or challenge and only inspect videos after the first pass.

The practical goal is to turn TikTok discovery into structured steps: find candidate videos or creators, enrich the records that matter, inspect comments when reaction quality matters, and stop before the workflow spends budget on low-value data.

TikTok workflows move between several objects: videos, creators, comments, hashtags, sounds, and playlists. If your application treats every task as “get more videos,” it misses the reason the lookup exists.

• Creator research starts with `user-info` and often follows with `user-videos`.
• Trend research starts with `search-videos`, `challenge-search`, or `music-videos` depending on the signal.
• Reaction analysis starts from a selected video and then calls `comment-list`.
• Media extraction starts from `video-info` when the workflow needs play URLs, covers, music metadata, and detailed counts.

This is why endpoint boundaries matter. A runtime can call the narrow API it needs instead of forcing a model to parse a rendered page and infer what object it is looking at.

For topic discovery, start with TikTok Search Videos. The required parameter is `keywords`, and the optional controls are where most workflow quality comes from: `count`, `cursor`, `region`, `publish_time`, and `sort_type`.

A trend monitor might search the same phrase by relevance first, then repeat with `sort_type=1` for like count or `sort_type=3` for date posted. A freshness-sensitive workflow can use `publish_time=1` for the past 24 hours, `7` for the current week, or broader windows when volume is low.

The response exposes video objects with IDs, captions, covers, music metadata, engagement counts, creation time, and author details. That is enough for a first-pass ranking step before you decide whether to inspect a creator, a comment thread, a sound, or the full video record.

Once a search result points to a creator worth inspecting, use TikTok User Info. It accepts `unique_id`, `user_id`, or both. The documented response includes profile fields such as handle, display name, avatar URLs, bio text, verification status, privacy flags, linked social handles, and stats like follower count, following count, total likes, and video count.

That gives your pipeline a stable creator record before it pulls more content. If the creator is relevant, follow with TikTok User Videos. This endpoint accepts `unique_id` or `user_id`, plus `count` and `cursor`, and returns posted videos with engagement fields and pagination state.

This sequence is cleaner than scraping a profile page in one pass. The profile call answers “who is this creator?” The videos call answers “what have they published recently enough to matter?”

For selected videos, use TikTok Video Info. The endpoint accepts a TikTok video URL or video ID in `url`, plus `hd=1` when the workflow needs the HD play URL.

The documented response includes video ID fields, region, title, cover URLs, duration, play URLs, music metadata, engagement counts, creation time, and author data. That makes it the right enrichment step after search or user-video discovery has already narrowed the candidate set.

For a one-off media extraction check before building the automated version, use the free TikTok video downloader workflow. It shows how the browser tool maps to the same `video-info` response fields and download links.

The rule is simple: search broadly, rank locally, enrich selectively. If you enrich every item before ranking, you make the workflow slower and harder to govern.

Engagement counts tell you that people reacted. They do not tell you why. For that, use TikTok Comment List. It accepts a video URL or video ID in `url`, plus optional `count` and `cursor` parameters.

Comment retrieval is useful for sentiment checks, objection mining, creator-audience fit, and detecting whether engagement is substantive or mostly noise. The response includes comment IDs, text, creation time, like counts, reply totals, status, and comment author data.

For agent workflows, comments should usually be a second-stage call. Let the model or ranking code choose which videos deserve the extra context.

TikTok hashtag workflows map to challenge endpoints in the docs. Use Challenge Search when you need to find relevant hashtags by keyword. It accepts `keywords`, with optional `count` and `cursor`, and returns challenge records with fields such as ID, name, description, user count, view count, type, and cover.

Once you have a challenge ID, use Challenge Posts to retrieve videos for that hashtag. It accepts `challenge_id`, plus optional `count`, `cursor`, and `region`.

This gives you a clean two-step hashtag workflow: resolve the challenge, then inspect the videos under it. That is easier to monitor than repeatedly searching keyword strings and hoping they map to the same trend surface.

Not every TikTok trend is best understood through hashtags. Some patterns spread through a sound. In that case, begin with Music Info when you need the sound record. It accepts a music ID or short URL and returns fields such as music ID, title, play URL, cover, author, original flag, duration, album, and video count.

Then use Music Videos to retrieve videos that use that sound. The required parameters are `music_id`, `count`, and `cursor`.

This branch is useful when an agent or analyst notices that a sound is driving repeated formats across creators. The workflow can pivot from one video to its music metadata, then to other videos using the same sound.

• Start with `search-videos` for the topic or phrase you care about.
• Rank returned videos by recency, engagement, region, author, or your own task-specific criteria.
• Use `user-info` for creators that appear relevant.
• Use `user-videos` to inspect recent creator output before overvaluing one viral post.
• Use `video-info` for selected videos that need full metadata or media URLs.
• Use `comment-list` only when audience reaction quality matters.
• Branch into `challenge-search`, `challenge-posts`, `music-info`, or `music-videos` when the trend is hashtag-led or sound-led.

This is the same shape as other structured retrieval workflows: discover, shortlist, enrich, then branch. The point is to keep each call tied to a decision instead of treating the API as an endless crawler.

TikTok research is a good fit for request-based access because the workflow is conditional. An agent may search one keyword and stop. It may inspect a creator after a promising result. It may pull comments only when engagement counts suggest a video is worth deeper review.

MintAPI exposes the same TikTok endpoint surface to human API key users and agent runtimes. Agent implementations should follow the request flow documentation: call the endpoint, handle `402 Payment Required` when returned, choose an accepted route, sign, and retry with `X-PAYMENT`.

For most implementations, the paidFetch and paidJson pattern is the cleaner starting point because payment handling stays in runtime code, not prompt instructions.

TikTok is different from text-first social monitoring because the signal may live in the video, the caption, the sound, the comments, the creator, or the hashtag. A good pipeline keeps those branches separate until there is a reason to join them.

If you are building a broader social intelligence system, compare this workflow with Twitter social signal workflows, YouTube transcript pipelines, Instagram API social listening workflows, and Facebook Marketplace parsing. The details change, but the discipline stays the same: start with a narrow endpoint, return structured JSON, and let the runtime decide whether deeper retrieval is worth the next call.

• Starting with comments before you know which videos are worth inspecting.
• Treating a single viral post as a creator profile without checking `user-info` or `user-videos`.
• Mixing keyword search, hashtag search, and music-led discovery into one vague crawler.
• Ignoring `cursor` and accidentally treating one page as complete coverage.
• Putting payment retries or signer choices inside model prompts.

The implementation lessons overlap with request-based API payments for agents and x402 vs API keys for agents. To orchestrate this kind of research in a visual automation workflow, see MintAPI for n8n. For a broader view of social media APIs across platforms, read API for social media workflows. Keep the model focused on tool choice and interpretation. Keep payments, retries, and request validation in deterministic code.

If the task is creator research, start with User Info and User Videos. If the task is trend discovery, start with Search Videos, then branch into hashtags, music, video details, or comments only when the first pass shows a reason to continue.