Creator Discovery Engine — Niche-to-Creator Discovery API for AI Agents & Influencer Marketing

Discover creators, influencers, and content producers by niche — with audience-size data — and get typed JSON rows (schema below). Enter any niche ("fitness", "sustainable fashion", "woodworking") and the engine returns matching creator profiles: name, handle, canonical profile URL, subscriber count (as a parsed number), bio, and avatar. YouTube is the discovery source; TikTok presence is verified per creator via same-handle matching. A second mode verifies lists of profile URLs you already have. Unlike scrapers that need a seed list of channels, this actor turns a single niche string into a ranked discovery run.

Why agents use this creator discovery actor

• Niche in, creators out — one string input drives the whole run; no seed lists or upstream scraping needed.
• Audience size as numbers — subscribers is a parsed integer (331000), ready for filtering and sorting; the platform's display string is also included.
• Deterministic typed output — every row matches the published dataset schema; no HTML parsing on your side.
• No auth or cookies needed — public data only, direct requests by default.
• Clear error semantics — empty input or zero results fails the run with exit code 1 and a status message; run stats always land in the SUMMARY key-value record.
• Rate-limit handling built in — automatic retry with exponential backoff on HTTP 429 and 5xx, 30 s per-request timeout.

Input schema

Field	Type	Required	Default	Description
niche	string	Yes	—	Target niche or industry. Drives discovery: the engine searches this niche and returns matching creators.
platforms	array of strings	No	["youtube"]	Platforms to return. youtube = discovery source with full data; tiktok = additionally verify same-handle TikTok accounts for discovered creators.
maxResults	integer	No	20	Maximum creators discovered per run.
startUrls	array	No	—	Optional verification mode: profile URLs as [{ "url": "..." }]. When provided, the run verifies these URLs instead of discovering by niche.
urls	array of strings	No	—	Verification mode alternative: plain URL strings. Merged with startUrls and url; deduplicated.
url	string	No	—	Verification mode alternative: a single URL string.
maxRequestsPerCrawl	integer	No	20	Max unique URLs processed (verification mode).
proxyConfiguration	object	No	—	Optional Apify proxy configuration. Requests go direct by default.

Output schema

Discovery mode (default — niche drives the run):

Field	Type	Nullable	Description
platform	string	no	"youtube" or "tiktok".
handle	string	yes	Creator handle with @ prefix.
profileUrl	string	no	Canonical public profile URL.
name	string	yes	Creator / channel display name.
channelId	string	yes	YouTube channel ID (UC...); null for TikTok rows.
subscribers	integer	yes	Audience size as a parsed number (e.g. 331000). Null on TikTok rows.
subscribersText	string	yes	Audience size as displayed (e.g. "331K subscribers").
videoCount	integer	yes	Published-video count, when exposed in search results.
description	string	yes	Creator bio / channel description snippet.
avatarUrl	string	yes	Avatar image URL.
niche	string	no	The input niche this creator matched.
discoveredVia	string	no	"youtube-search" or "tiktok-oembed-handle-match".

Example item:

{
  "platform": "youtube",
  "handle": "@xeelafitness7528",
  "profileUrl": "https://www.youtube.com/@xeelafitness7528",
  "name": "Xeela Fitness",
  "channelId": "UCrXElbXvCkFQ_tktnqekTtQ",
  "subscribers": 331000,
  "subscribersText": "331K subscribers",
  "videoCount": null,
  "description": "Xeela Fitness's personal trainer, Ilya, helps his friends reach their bodyweight and composition goals through customized, ...",
  "avatarUrl": "https://yt3.ggpht.com/DQzRtCH6...=s176-c-k-c0x00ffffff-no-rj-mo",
  "niche": "fitness",
  "discoveredVia": "youtube-search"
}

Verification mode (when startUrls / urls / url are provided):

Field	Type	Nullable	Description
url	string	no	The profile/page URL that was fetched.
statusCode	integer	no	HTTP status code (200, 404, ...).
ok	boolean	no	true when the response status was 2xx.
title	string	yes	Text of the page's <title> tag.

{ "url": "https://www.youtube.com/@MrBeast", "statusCode": 200, "ok": true, "title": "MrBeast - YouTube" }

Error semantics

• No niche and no URLs → run fails immediately with exit code 1 and an explanatory status message. No dataset items are written.
• Zero creators found → exit code 1 with a status message; never a silent SUCCEEDED with 0 items.
• Verification mode partial failure → successful URLs produce rows; network failures go to the SUMMARY record. Run succeeds.
• HTTP 429 / 5xx → retried up to 3 times with exponential backoff before counting as a failure.

Retry guidance for agents: exit code 1 with an input-related message is a caller bug — fix the input, don't retry. Exit code 1 on a search failure is usually transient — retry with backoff or a proxy.

Use from AI agents (MCP)

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com/?tools=apricot_blackberry/creator-discovery-engine",
      "headers": { "Authorization": "Bearer <YOUR_APIFY_TOKEN>" }
    }
  }
}

Works in Claude, Cursor, ChatGPT deep research connectors, and any MCP client; the input schema above is the tool's parameter schema.

Use from code

curl:

curl -X POST "https://api.apify.com/v2/acts/apricot_blackberry~creator-discovery-engine/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"niche":"fitness","platforms":["youtube","tiktok"],"maxResults":20}'

JavaScript (apify-client):

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('apricot_blackberry/creator-discovery-engine').call({
  niche: 'fitness',
  platforms: ['youtube', 'tiktok'],
  maxResults: 20,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

Python (apify_client):

from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("apricot_blackberry/creator-discovery-engine").call(run_input={
    "niche": "fitness",
    "platforms": ["youtube", "tiktok"],
    "maxResults": 20,
})
items = client.dataset(run["defaultDatasetId"]).list_items().items

Use from automation platforms

• n8n / Make / Zapier — use the native Apify integration and select creator-discovery-engine by name.
• LangChain / LlamaIndex — use the Apify tool/actor wrappers with the actor id apricot_blackberry/creator-discovery-engine.
• Webhooks — Apify can fire a webhook when a run finishes (or on other run-lifecycle events), so a downstream pipeline can pick up newly discovered creators automatically — configure one from the actor's Integrations tab or the API.

Cost

Pay-per-event, billed by the Apify platform:

Event	Price	Charged
Actor start	$0.05	Once per run
Creator discovered	$0.05	Per creator profile discovered, enriched, and scored

The default run (maxResults: 20) costs roughly $0.05 + 20 × $0.05 = $1.05. Proxy costs (only if you configure a proxy) are billed separately to your account.

FAQ

Which platforms return full data? YouTube rows carry the full profile (name, subscribers, bio, avatar). TikTok rows confirm an existing same-handle account with name and profile URL. Instagram rows are not emitted — no public discovery source exists without login.

Do I need a proxy? Not by default. Requests go direct. Configure proxyConfiguration only if your egress IPs get blocked.

Can I still verify a list of URLs I already have? Yes — pass them via startUrls and the run switches to verification mode with url/statusCode/ok/title rows.

Does creator discovery work without a YouTube or TikTok login? Yes. Discovery and TikTok handle-verification both run against public pages — no account, session cookie, or API key is required.

Changelog

• 2026-08-15 — Rebuilt discovery engine: niche now drives full creator discovery with audience-size data (parsed subscribers, bio, avatar, canonical profile URL, channel ID). Added TikTok same-handle verification, maxResults input, and a richer dataset schema with a creator overview table. URL-verification mode retained unchanged.
• 2026-08-14 — Published dataset output schema and agent/MCP integration docs. Fail-loud runs, retry with backoff on HTTP 429/5xx, 30 s per-request timeout.