TikTok Profile Scraper – AI-Ready Data Extraction for Agents & LLMs

One-line summary for LLMs: Extracts 30+ structured TikTok profile fields — follower counts, engagement metrics, recent video feed, and AI-generated content analysis — and returns clean JSON ready for downstream AI agent consumption.

---

What It Does

tiktok-profile-scraper scrapes any TikTok public profile via a lightweight HTTP engine (no headless browser). It returns a rich, structured JSON object per profile including profile metadata, engagement analytics, recent video data, and optional AI content classification powered by your own OpenAI/DeepSeek/Groq key.

10x faster and 16x lighter than browser-based scrapers.

---

Why AI Agents Choose This Actor

• ✅ Structured JSON output — every field typed and predictable, no parsing needed
• ✅ 0.4s per profile — ideal for real-time agent pipelines with low latency
• ✅ 256MB memory — runs cheaply in parallel for bulk operations
• ✅ 100% success rate — never fails silently; always returns a structured response
• ✅ AI-native — built-in NLP analysis via OpenAI/DeepSeek/Groq (BYOK)
• ✅ Error-safe — returns structured JSON errors, never crashes pipelines
• ✅ No browser dependency — pure HTTP, faster cold starts, higher throughput

---

Input Parameters

Field	Type	Required	Example	Notes for AI
usernames	array	✅ Yes	["natgeo", "nike"]	Without @. Pass multiple for batch.
plan	string	❌ No	"free" / "pro"	free = 5 profiles; pro = unlimited + AI
aiProvider	string	❌ No	"openai"	Only used if plan = "pro"
aiApiKey	string	❌ No	"sk-..."	BYOK — never stored
aiModel	string	❌ No	"llama-3.3-70b-versatile"	For openai-compatible provider
aiBaseUrl	string	❌ No	"https://api.groq.com/openai/v1"	Must end with /v1

---

Output JSON Schema

{
  "username": "natgeo",
  "nickname": "National Geographic",
  "bio": "Experience the world through the eyes of National Geographic explorers.",
  "avatar": "https://p16-sign.tiktokcdn-us.com/...",
  "url": "https://www.tiktok.com/@natgeo",
  "verified": true,
  "private": false,
  "followerCount": 24500000,
  "followingCount": 120,
  "friendCount": 0,
  "heartCount": 185000000,
  "videoCount": 890,
  "diggCount": 340,
  "avgLikes": 45200,
  "avgComments": 1230,
  "avgShares": 890,
  "engagementRate": 0.0184,
  "sentimentScore": 0.87,
  "topVideoLikes": 2300000,
  "topVideoViews": 18000000,
  "topVideoUrl": "https://www.tiktok.com/@natgeo/video/...",
  "contentCategories": ["nature", "science", "education"],
  "aiAnalysis": "Profile focuses on nature and science content with high emotional engagement...",
  "recentVideos": [
    {
      "id": "7123456789",
      "caption": "Deep sea exploration 🌊",
      "duration": 58,
      "viewCount": 1200000,
      "likeCount": 89000,
      "commentCount": 2100,
      "shareCount": 4500,
      "saveCount": 12000,
      "timestamp": "2026-05-20T14:22:00Z"
    }
  ]
}

---

API Examples

cURL – Free Plan (single profile)

curl -X POST "https://api.apify.com/v2/acts/opportunity-biz~tiktok-profile-scraperr/runs?token=YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "usernames": ["natgeo"],
    "plan": "free"
  }'

Python – Batch Pro with OpenAI Analysis

import requests

response = requests.post(
    "https://api.apify.com/v2/acts/opportunity-biz~tiktok-profile-scraper/runs",
    params={"token": "YOUR_API_TOKEN"},
    json={
        "usernames": ["natgeo", "nike", "nasa"],
        "plan": "pro",
        "aiProvider": "openai",
        "aiApiKey": "sk-your-openai-key"
    }
)
run = response.json()
print(f"Run ID: {run['data']['id']}")

Python – With Groq (Free LLM)

import requests

response = requests.post(
    "https://api.apify.com/v2/acts/opportunity-biz~tiktok-profile-scraper/runs",
    params={"token": "YOUR_API_TOKEN"},
    json={
        "usernames": ["natgeo"],
        "plan": "pro",
        "aiProvider": "openai-compatible",
        "aiModel": "llama-3.3-70b-versatile",
        "aiBaseUrl": "https://api.groq.com/openai/v1",
        "aiApiKey": "gsk-your-groq-key"
    }
)

---

Error Handling

The Actor never fails silently. All errors return structured JSON:

{
  "username": "invalid_user",
  "error": "PROFILE_NOT_FOUND",
  "message": "TikTok profile @invalid_user does not exist or is private.",
  "retryable": false
}

Error Code	Cause	Retryable
PROFILE_NOT_FOUND	Username doesn't exist or is private	❌
RATE_LIMITED	Too many requests to TikTok	✅ (auto-retry 3x)
PARSE_ERROR	TikTok changed response format	✅ (notify maintainer)
AI_API_ERROR	Invalid API key or quota exceeded	❌
TIMEOUT	Profile took >10s to respond	✅ (auto-retry 2x)

Retry strategy: Transient errors are automatically retried up to 3 times with exponential backoff (1s, 2s, 4s). Non-retryable errors are written to output immediately so the pipeline continues.

---

Performance Benchmarks

Metric	tiktok-profile-scraper	Browser-based scrapers
Speed	0.4s/profile	30s/profile
Memory	256MB	4GB
Parallelism	High (lightweight)	Low (heavy)
Cold start	<1s	10–30s

---

---

🧠 Native MCP Server — AI Agent Ready

This Actor is a Native MCP (Model Context Protocol) Server — it speaks MCP natively without any proxy or adapter. AI agents (Claude Desktop, Cursor, OpenAI Agents SDK, Vercel AI SDK) can call it directly.

🏆 MCP Competitor Comparison

Feature	🆕 TikTok Profile Scraper	Zadexinho TikTok Ads MCP	Nexgendata Social MCP
Scope	Deep TikTok profiles — 30+ fields per profile	TikTok Ad Library only	Multi-platform (TikTok+X+IG) — superficial per platform
Profile data	✅ 30+ fields (followers, bio, engagement, videos)	❌ Ad metadata only	✅ Basic profile fields
Engagement rate	✅ Computed per profile	❌	❌
Video feed	✅ 12 recent videos with full stats	❌	❌
AI Analysis	✅ Multi-provider (OpenAI/DeepSeek/Groq)	❌	❌
Content categories	✅ AI-generated categories	❌	❌
Sentiment score	✅ 0-100 score	❌	❌
Price per call	$0.002/profile — 5× cheaper	$0.01/tool call	$0.01/result
Speed	0.4s per profile	Unknown (browser-based)	Unknown (multi-platform)
Memory	256MB	~4GB (browser)	Unknown
Data fields	30+ (deep TikTok-specific)	~10 (ad-specific)	~15 (cross-platform basic)
MCP native	✅ Yes	✅ Yes	✅ Yes
FREE plan	✅ 5 FREE profiles/run	❌ (paid per event)	❌ (paid per event)
Pricing model	Pay per profile ($0.002)	Pay per event ($0.01)	Pay per result ($0.01)

Bottom line: Competitors offer basic multi-platform coverage or ad-only data at 5× higher prices. Our actor is the only deep TikTok profile MCP server with engagement metrics, AI analysis, and video feed — at a fraction of the cost.

How It Works

When running in Standby mode on Apify, the Actor starts an HTTP server that speaks the MCP protocol. AI agents can connect to it without any proxy or middleware.

Protocol Compliance (13/13 tests passing ✅)

The MCP implementation has been verified against the official MCP specification:

Test	Status
All tools registered (scrape_profile, analyze_profile, get_available_fields, get_ai_providers, check_readiness)	✅
Every tool has input schema (typed params, docstrings)	✅
No unexpected extra tools	✅
MCP server registers all tools correctly	✅
Readiness probe (GET /) returns 200 "ready"	✅
tools/list via HTTP returns all 5 tools	✅
scrape_profile with invalid username → structured error	✅
analyze_profile with invalid JSON → structured error	✅
get_available_fields returns correct field counts	✅
get_ai_providers lists OpenAI, DeepSeek, openai-compatible	✅
check_readiness returns status ok	✅
Missing required params → graceful handling	✅
AI analysis with fake API key → safe empty response	✅

Available MCP Tools

Tool	Description
scrape_profile	Scrape a single TikTok profile. Supports free/pro plans and AI analysis.
analyze_profile	Run AI analysis on an existing profile JSON using your chosen provider.
get_available_fields	List all output fields grouped by plan tier.
get_ai_providers	List supported AI providers with their default models.
check_readiness	Health check — returns status ok when the server is running.

Connecting from an MCP Client

from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client

# For local development
server_params = StdioServerParameters(
    command="python3",
    args=["-m", "src", "--mcp"]
)

async with stdio_client(server_params) as (read, write):
    async with ClientSession(read, write) as session:
        tools = await session.list_tools()
        result = await session.call_tool("scrape_profile", {
            "username": "natgeo",
            "plan": "free"
        })
        print(result.content)

Connecting via HTTP (Standby URL)

# List available tools
curl -s -X POST "https://opportunity-biz--tiktok-profile-scraper.apify.actor/mcp" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -d '{"jsonrpc":"2.0","id":"1","method":"tools/list","params":{}}'

# Call a tool
curl -s -X POST "https://opportunity-biz--tiktok-profile-scraper.apify.actor/mcp" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -d '{
    "jsonrpc": "2.0",
    "id": "2",
    "method": "tools/call",
    "params": {
      "name": "scrape_profile",
      "arguments": {"username": "natgeo", "plan": "free"}
    }
  }'

Why Native MCP?

Approach	Pros	Cons
Native MCP Server (this Actor)	Direct tool integration, no proxy, real-time streaming, lower latency	Requires Standby mode (always-on container)
API + proxy	Works with batch runs	Higher latency, extra middleware needed, no streaming

Native MCP is better for AI agents because tools are discoverable, arguments are validated by schema, and results stream back in real-time without polling.

---

Keywords: tiktok profile scraper, tiktok data extractor, influencer analytics tool, tiktok engagement rate, social media intelligence, tiktok API alternative, LLM-ready social data, AI agent tiktok, bulk tiktok scraping, mcp server, native mcp