TikTok MCP Server已验证已精选

TikTok creator, video and comment research for AI agents

TikTok MCP Server gives Claude, Cursor, Windsurf and other MCP clients access to public TikTok profiles, account videos, comments and keyword search. Maintained by HasData, it returns structured JSON for creator research, content comparisons and audience feedback analysis. The server supplies the data; your AI agent calculates rankings, compares metrics and summarizes comments. Repository and setup guide.

HasData hosts the server. You need an MCP client and a HasData account, with no TikTok developer application or TikTok login required.

Prompts to try

• Creator discovery: “Search for TikTok creators matching ‘specialty coffee’. Compare the returned accounts by follower count and bio, and explain which appear relevant to a coffee equipment brand.”
• Content comparison: “Read @nasa’s profile and the first two pages of its videos. Calculate median plays across the returned videos and link the five with the most shares.”
• Audience feedback: “Read the first page of comments on this TikTok video, then fetch replies to the comment with the most likes in that page. Summarize recurring questions and distinguish the sample from the full discussion.”

These are research prompts, not saved reports. Results and metrics depend on the data returned by each call.

Four available tools

Tool	Parameters and returned data
hasdata_tiktok_profile_getTikTokProfile	Required handle, with or without @. Returns profile details such as bio, verification status, follower count, total likes and video count.
hasdata_tiktok_posts_getTikTokPosts	Required handle; optional nextPageToken. Returns a page of account videos with descriptions, URLs, timestamps and engagement counters including plays, likes, comments and shares.
hasdata_tiktok_comments_getTikTokComments	Required videoId; optional commentId and nextPageToken. Returns comment text, author details, likes, timestamps and reply counts. Supply commentId to retrieve replies.
hasdata_tiktok_search_getTikTokSearch	Required keyword; optional type and nextPageToken. type defaults to video; use user for creator results with handles, bios and follower counts.

The tool reference documents every parameter. Posts, comments and search support pagination: pass the previous response’s nextPageToken for another page. Keep video and comment IDs as strings. Optional fields, including hashtags or a bio link, may be absent.

Connect

Use streamable HTTP at https://mcp.hasdata.com/api/mcp?apis=tiktok with your HasData API key in the x-api-key header. For Cursor:

{
  "mcpServers": {
    "tiktok": {
      "url": "https://mcp.hasdata.com/api/mcp?apis=tiktok",
      "headers": {
        "x-api-key": "YOUR_HASDATA_API_KEY"
      }
    }
  }
}

Create a key in the HasData dashboard. Compatible OAuth clients can instead connect and sign in with HasData. Clients requiring stdio can use the @hasdata/tiktok-mcp npm launcher or hasdata-tiktok-mcp Python launcher. See client-specific setup.

Free allowance and cost

Each successful TikTok tool call costs 10 credits. The free plan provides 1,000 credits every month, with no credit card required: 100 TikTok calls if the allowance is used entirely here. A profile lookup plus two pages of videos uses three calls, or 30 credits. Additional pages and reply requests are separate calls. See current plans and limits.

Common questions

Is this an official TikTok service?

No. HasData maintains this independent service for reading publicly accessible TikTok data. It does not provide private account access, posting or direct messages.

Can I retrieve one video’s details or transcript by URL?

There is no dedicated video-detail or transcript tool in this server. Obtain video metadata through account posts or keyword search. For comments, extract the numeric ID after /video/ from the URL.

Where can I inspect the underlying API?

The TikTok Scraper API page includes search and account-post request builders. The TikTok Profile API page covers profile lookups. These show the underlying APIs; MCP connection instructions are in the MCP documentation.