For AI

Answers

How do I install Xiaohongshu, Douyin and WeChat data skills in OpenClaw?

Short answer

Run `openclaw skills install @conn-ho/okdata-xiaohongshu` (and okdata-douyin, okdata-wechat-mp), put your okdata key in ~/.openclaw/.env as OKDATA_API_KEY, restart the Gateway, then ask in plain words. One page of Xiaohongshu note search is $0.02 for about 20 notes, the Douyin hot search board is $0.002, and a realistic first week of 46 calls comes to $0.704.

An OpenClaw user filed a request on GitHub: when the agent gets a Douyin or Xiaohongshu link, it opens a browser by default, and he has to add "use the skill" every time. That only helps if a dependable data skill is installed. okdata publishes one each for Xiaohongshu, Douyin and WeChat Official Accounts on ClawHub. They call an API for public data, so no platform account, cookie or QR login is involved.

Set it up once

Install the skills. OpenClaw's docs give openclaw skills install @owner/<slug>, and okdata's skills are under conn-ho:

openclaw skills install @conn-ho/okdata-xiaohongshu
openclaw skills install @conn-ho/okdata-douyin
openclaw skills install @conn-ho/okdata-wechat-mp
openclaw skills install @conn-ho/okdata

The last one is the general skill: it lets the agent pick from all 265 endpoints on 30 platforms, TikTok, YouTube, Instagram, X, Reddit and Amazon among them. Skills go into the workspace skills/ folder; --global puts them in ~/.openclaw/skills for every local agent. If your version rejects the @conn-ho/ prefix, drop it. We checked every skill named here with clawhub inspect on November 18; ClawHub's scan verdict was clean for all of them.

Then get a key in the console (a new key comes with $0.10 of credit), add OKDATA_API_KEY=okd_... to ~/.openclaw/.env, and run openclaw gateway restart. OpenClaw reads that file at startup without overriding variables already set. If the agent runs in a sandbox, make sure the variable reaches it.

Last, in a new session, ask: "Follow https://okdata.online/agents.md to check that okdata is connected; the key is in OKDATA_API_KEY. Make no paid call; tell me the balance." agents.md is a four-step checklist the agent works through.

Then ask in plain words

Name the skill or say "use okdata", so the agent does not reach for the browser first.

You say Endpoint Cost
"Use okdata-xiaohongshu to search 露营 装备, most popular, one page; list the 5 most-liked notes" Xiaohongshu note search $0.02
"What are people saying under this note? Read two pages of comments" Xiaohongshu note comments, about 10 a page $0.04
"Show today's top 20 on Douyin hot search" Douyin hot search board, about 50 entries $0.002
"List the newest page of articles from Official Account gh_xxxx and read two in full" Official Account article history plus 2 × article detail $0.06

Search Chinese platforms in Chinese; the notes and videos are written in it.

A first week, priced

A small outdoor-gear team checks Douyin hot search every morning, searches a few keywords on Xiaohongshu and Douyin, reads comments, and follows three rivals' Official Accounts:

Job Calls Cost
Douyin hot search, daily 7 $0.014
Xiaohongshu: 5 keywords × 2 pages, then comments on 10 notes 20 $0.40
Douyin: 3 keyword searches, then comments on 5 videos 8 $0.07
WeChat: find 3 accounts by name (live search), newest page each, read 5 articles 11 $0.22
The week 46 $0.704

Pay per call, no subscription; every endpoint has its own price on the pricing page. Your model's own tokens are separate.

The same calls in Python

import os, requests

API = "https://api.okdata.online"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}

def get(path, **params):
    body = requests.get(API + path, headers=H, params=params, timeout=60).json()
    if body["code"] != "ok":
        raise RuntimeError(f"{path}: {body['code']} {body['message']}")
    return body["data"]["items"]

for t in get("/v1/douyin/trends/list", fields="rank,name,heat")[:10]:
    print(t["rank"], t["name"], t["heat"])

notes = get("/v1/xiaohongshu/notes/search", query="露营 装备", sort="popular", fields="id,url,title,metrics.likes")
top = max(notes, key=lambda n: n["metrics"]["likes"] or 0)
for c in get("/v1/xiaohongshu/notes/comments/list", id=top["id"], fields="text,likes"):
    print(c["likes"], c["text"])

MCP instead of, or next to, the skills

OpenClaw can save a remote MCP server: openclaw mcp set okdata '{"url":"https://api.okdata.online/mcp","transport":"streamable-http","headers":{"Authorization":"Bearer ${OKDATA_API_KEY}"}}'. Keep the single quotes so the reference is stored and resolved from the environment. The skills tell the agent which endpoint answers which question and what it costs; with MCP connected it calls through MCP, otherwise over HTTP. See the MCP page.

Not on OpenClaw? npx skills add okdataapi/skills installs the set into Claude Code, Codex and other agents that read SKILL.md. Related: writing your own Claude Code skill, WorkBuddy, Doubao and Qwen.

Sources

Where people asked this (read 2026-11-18)

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