Use Douyin, RedNote and Weibo data in Claude Code and Cursor with MCP

You can ask an AI assistant "what are people saying about this product on Xiaohongshu?" and have it actually look. MCP (Model Context Protocol) lets Claude Code, Cursor, VS Code and other AI tools call outside services. Qovia runs an MCP server for its Chinese and global social and e-commerce data: once it is connected, the AI finds the right endpoint, pulls the data and sums it up for you.

What you get

• Server: https://api.qovia.ai/mcp, over Streamable HTTP, authenticated with the same API key as the REST API.
• Four tools by default: search_endpoints (find an endpoint), get_endpoint (read its parameters), call_endpoint (run it) and get_account (check the balance). The full catalog stays out of the AI's context until it needs it.
• The same data as the API: Douyin, Xiaohongshu (RedNote), Weibo, WeChat official accounts, Taobao and JD reviews and more. See the platform directory.

Step 1: create a key for the AI

Create an API key in the console just for your AI tools, and tick only the scopes it needs, such as raw:douyin or raw:xiaohongshu. You can revoke it at any time without touching your other integrations.

Step 2: connect your client

Claude Code

Run this once in a terminal, with your key in QOVIA_KEY:

claude mcp add --transport http qovia https://api.qovia.ai/mcp \
  --header "Authorization: Bearer $QOVIA_KEY"

Cursor

Add Qovia to ~/.cursor/mcp.json, with your own key, then restart Cursor:

{
  "mcpServers": {
    "qovia": {
      "url": "https://api.qovia.ai/mcp",
      "headers": { "Authorization": "Bearer sk-qovia-..." }
    }
  }
}

VS Code

In .vscode/mcp.json, VS Code asks for the key on first use, so it is never written to the file:

{
  "servers": {
    "qovia": {
      "type": "http",
      "url": "https://api.qovia.ai/mcp",
      "headers": { "Authorization": "Bearer ${input:qovia-key}" }
    }
  },
  "inputs": [{ "type": "promptString", "id": "qovia-key", "description": "Qovia API key", "password": true }]
}

Claude Desktop and other clients are covered in the MCP docs.

Step 3: ask in plain language

Some questions to start with:

• "Search Xiaohongshu for 露营 (camping), sorted by most liked, and summarise what the top 20 notes have in common."
• "Pull the first 100 comments on Douyin video 7681194457378688292. What do people praise and what do they complain about?"
• "What are the top 10 on Weibo's hot search right now? Group them by topic."

Behind the scenes the AI works step by step. For the Douyin question it searches the catalog (search_endpoints), reads video comments (get_endpoint), then calls it with the video ID and a page size (call_endpoint), paging with the cursor until it has enough.

Optional: give the AI the endpoints directly

Add ?tools= to the server address to expose endpoints as tools of their own, for every platform (?tools=all) or only some:

https://api.qovia.ai/mcp?tools=xiaohongshu,douyin

Each endpoint becomes a tool named after its id, with dots as underscores (raw.douyin.comments becomes raw_douyin_comments). The four default tools stay.

Costs and safety

• MCP calls are billed exactly like API calls: only successful calls that return data are charged, at each endpoint's price (see pricing). Empty results and errors are not. Finding and reading endpoints costs nothing, and neither does get_account.
• Tell the AI how much to fetch ("the first 100 comments", "3 pages") to keep a question's cost predictable.
• The per-minute limit is per account and shared with your API calls. When the AI hits it, the tool error says how long to wait.

Next

• Prefer code? Get Xiaohongshu notes and comments with Python or export Douyin comments to Excel.
• Every endpoint, with its parameters and a real response, is in the API reference.