For AI

Answers

How can Doubao read Douyin and Xiaohongshu data?

Short answer

On its own, Doubao can show Douyin search results and summarize the page you are viewing, but its official docs describe no way to pull comments or Xiaohongshu notes in bulk. Add https://api.okdata.online/mcp with your key as a custom connector in Doubao Work and it can search Douyin videos ($0.02 a page), read Douyin comments ($0.002 a page) and search Xiaohongshu notes ($0.02 a page). Reading this week's comments on a product's ten top Douyin videos plus five Xiaohongshu notes costs about $0.50.

An e-commerce operator keeps Doubao open all day. Before Monday's meeting she wants to know what people complained about this week under Douyin videos of her air fryer. Asked directly, Doubao writes a plausible summary, but there is no telling which videos or comments it looked at, and the counts cannot be checked. To make it work from real comments, the comments have to reach it.

What Doubao does on its own

According to Doubao's own documentation:

  • With a Douyin account linked, Doubao can show Douyin search results in the chat (Doubao privacy policy).
  • The desktop app and browser extension summarize and translate the page you are viewing, including video sites (extension page and privacy policy).
  • Doubao Work installs plugins and lets you add your own connector for an internal API or an MCP server (Doubao Work help center, "使用插件").

That covers one video or one note at a time. Ten videos, a thousand comments, this week only, counted by theme: we found nothing in the docs for that, and nothing about searching Xiaohongshu notes or reading their comments.

Connect okdata in Doubao Work

In 插件·技能·伙伴, click + 添加 on the plugins page, then 新建自定义连接器. Choose HTTP, set the server URL to https://api.okdata.online/mcp, and add a header Authorization with the value Bearer okd_.... Save, start a new task and pick okdata with /. Asking for your balance is free and confirms the link. Full steps, and the Volcano Engine Ark route for developers (Responses API with remote MCP), are on the Doubao integration page.

Then ask: search Douyin for the product sorted by latest, three pages; keep the ten videos from the last seven days with the most comments; read five comment pages each; keep this week's comments; group the complaints with counts and quotes; estimate the cost first.

What that job costs

Step Calls Cost
Douyin search, 3 pages 3 $0.06
Douyin comments, 10 videos × 5 pages 50 $0.10
Xiaohongshu search, 2 pages 2 $0.04
Xiaohongshu comments, 5 notes × 3 pages 15 $0.30
Total 70 $0.50

The Douyin hot search is $0.002 a call. A call that reached the data source is charged even when it comes back empty, not found or failed; calls stopped earlier, such as bad parameters or low balance, are free, and find, describe, read_more and account cost nothing.

Without a connector

import os
import requests
from datetime import datetime, timedelta, timezone

API = "https://api.okdata.online"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}
SINCE = (datetime.now(timezone.utc) - timedelta(days=7)).strftime("%Y-%m-%dT%H:%M:%SZ")

def pages(path, n, **params):
    cursor = None
    for _ in range(n):
        q = {**params, "cursor": cursor} if cursor else params
        body = requests.get(API + path, headers=H, params=q, timeout=60).json()
        if body["code"] != "ok":
            print("skipped", path, body["code"], body["message"])
            return
        yield from body["data"]["items"]
        if not body["page"]["has_more"]:
            return
        cursor = body["page"]["next_cursor"]

videos = [v for v in pages("/v1/douyin/videos/search", 3, query="XX 空气炸锅", sort="latest",
                           fields="id,published_at,metrics.comments") if v["published_at"] >= SINCE]
top = sorted(videos, key=lambda v: v["metrics"]["comments"] or 0, reverse=True)[:10]
lines = [f"{c['likes'] or 0}\t{c['text']}"
         for v in top
         for c in pages("/v1/douyin/videos/comments/list", 5, id=v["id"], fields="text,likes,published_at")
         if c["published_at"] >= SINCE]
with open("comments.txt", "w", encoding="utf-8") as f:
    f.write("\n".join(lines))

Drop comments.txt into the Doubao chat and ask for complaint themes with counts. Douyin returns comments hottest first only, so the date filter is yours to apply. For a tag-and-count method see finding pain points in comments; for Coze workflows, Coze and Dify; for hot lists in a prompt, DeepSeek and hot search.

Sources

Where people asked this (read 2026-11-17)

Updated

Connect once, then just ask.

Every new key comes with $0.10 to try it. One sentence connects your agent.

Start free
Connect your AI

Paste this into your agent

Read https://okdata.online/agents.md and set up okdata for me; my key is in the OKDATA_API_KEY environment variable.

Claude Code, Cursor, Codex and other agents that can open a link set themselves up from it.

MCP serverhttps://api.okdata.online/mcp