TikTok Comments Scraper · Apify Actor

Extract every comment from any TikTok video — including reply threads, author info, like counts, pinned status, and timestamps. Sentiment-analysis and spam-detection ready.

$0.50 / 1,000 comments — cheapest in the category (head actor: $0.50–$0.70 / 1K). Pay-per-result, failed requests are free.

---

中文说明

输入一个或多个 TikTok 视频链接，抓取每条视频的全部评论（含回复线程、作者信息、点赞数、置顶状态、时间戳）。

工作原理（可靠的签名翻页）：用真实浏览器打开视频页，让 TikTok 自身先成功发出一次 API 请求以预热 msToken/会话；随后捕获或构造 comment/list 请求模板，每页只换 cursor，用页面内置签名函数 byted_acrawler.frontierSign 重新签名（X-Bogus），跟随 has_more 翻页。回复线程通过把模板改写为 comment/list/reply 接口（换 item_id+comment_id）同样签名翻页。默认使用 Apify Proxy DEFAULT 非住宅代理池，不会自动切到住宅代理；如果用户手动选择 RESIDENTIAL，则尊重该输入。

输入：videoUrls（视频链接，必填）、commentsPerVideo（每视频最多评论数，0=全部）、includeReplies（是否抓回复）、repliesPerComment、proxy（默认 Apify Proxy DEFAULT 非住宅池）、maxConcurrency、maxRequestRetries。 输出：每行一条评论（顶层或回复），含 commentId/text/author/diggCount/replyCount/isReply/parentCommentId/isPinned/createTimeISO 等。 计费：按结果 $0.50 / 1000 条；失败请求不收费。

---

Why this actor

Feature	This actor	Typical alternatives
Extraction surface	Signed API pagination — re-signs each /api/comment/list/* request with TikTok's own in-page signer	Reverse-engineering rotating signatures every week
Reply threads	Yes (configurable depth + per-thread cap)	Often top-level only
Output fields	18 (incl. isAuthorLiked, isPinned, mentionedUsers, depth)	8–12
MCP-compatible	Yes — discoverable via mcp.apify.com	Most are not
Pricing	$0.50 / 1K comments	$0.50–$0.70 / 1K
Proxy	Non-residential DEFAULT proxy by default; explicit RESIDENTIAL is respected	Often residential-only or datacenter-only

---

Input

{
  "videoUrls": [
    "https://www.tiktok.com/@khaby.lame/video/7321234567890123456",
    "https://www.tiktok.com/@charlidamelio/video/7345678901234567890"
  ],
  "commentsPerVideo": 200,
  "includeReplies": true,
  "repliesPerComment": 20,
  "proxy": {
    "useApifyProxy": true
  },
  "maxConcurrency": 2,
  "maxRequestRetries": 2
}

Full schema → ./.actor/input_schema.json.

---

Output (one row per comment)

{
  "commentId": "7398234567890",
  "videoId": "7321234567890123456",
  "videoUrl": "https://www.tiktok.com/@khaby.lame/video/7321234567890123456",
  "text": "lmao 💀 you nailed it",
  "createTimeISO": "2026-05-12T15:01:22.000Z",
  "isReply": false,
  "parentCommentId": null,
  "depth": 0,
  "diggCount": 4231,
  "replyCount": 12,
  "isAuthorLiked": true,
  "isPinned": false,
  "author": {
    "id": "6912345678",
    "uniqueId": "fan_name",
    "nickname": "Fan Name",
    "verified": false,
    "avatar": "https://...",
    "isVideoAuthor": false
  },
  "mentionedUsers": [],
  "scrapedAt": "2026-05-19T07:34:18.012Z"
}

Replies appear in the same dataset with isReply=true and parentCommentId populated. Full schema → ./.actor/dataset_schema.json.

---

Pricing — Pay per result

Event	Charge
comment-scraped	$0.50 / 1,000 comments (top-level OR reply)

Failed requests and retries are free.

---

How it works (extraction strategy)

1. Open the video page with Playwright using fingerprint randomisation + Apify Proxy DEFAULT non-residential pool by default.
2. Sniff XHR responses that match /api/comment/list/* — these are TikTok's own signed requests, so we never need to compute X-Bogus / msToken ourselves.
3. Scroll briefly or build a synthetic comment template when the page never fires its own comment request.
4. (Optional) click "View replies" affordances so reply-thread XHRs fire too.
5. De-dup → trim to limits → push → charge.

This is the same architecture used by the top-rated TikTok scrapers — robust to weekly signature rotations, since the page does the signing for us.

---

Local dev & deploy

npm install
mkdir -p storage/key_value_stores/default
cat > storage/key_value_stores/default/INPUT.json <<'JSON'
{
  "videoUrls": ["https://www.tiktok.com/@khaby.lame/video/7321234567890123456"],
  "commentsPerVideo": 50
}
JSON
npm start

Push to Apify:

npm install -g apify-cli
apify login
cd apify-tiktok-comments-scraper
apify push

After upload, set monetization to Pay per result → bind comment-scraped event → $0.50 / 1K.

---

Using this actor via MCP (Claude / Cursor / ChatGPT)

The actor is auto-exposed as a tool by Apify MCP Server at https://mcp.apify.com/.

Cursor / Claude Desktop config

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com/",
      "transport": "streamable-http",
      "headers": {
        "Authorization": "Bearer apify_api_xxxxxxxx"
      }
    }
  }
}

Then, in chat:

"Use the tiktok-comments-scraper actor to grab the 200 most-liked comments on https://www.tiktok.com/@khaby.lame/video/7321234567890123456 — include replies."

Apify MCP Server reads the input_schema.json to build the tool definition and the dataset_schema.json to advertise the output shape to the LLM, so the agent knows what fields to expect back.

---

License & disclaimer

MIT. Scrape only public data and respect TikTok's terms. You are responsible for compliance with applicable laws and platform policies.