hot-topics-mcp
Trending topics for AI agents — Weibo / Baidu / Zhihu (China) + Google Trends (US), plus HackerNews & StackExchange community intel. 2 tools. Zero dependencies, pure Python stdlib.
- Hosted (no install):
https://hot.mcp.goaimoat.com/mcp - Registry name:
com.goaimoat/hot-topics - Free: 10 calls per day per IP. Then $0.03/call.
Built for content-creation and marketing agents that need to know what people are actually talking about right now before they write. Battle-tested in production driving a daily automated content pipeline.
Tools
Why this exists
- Trending data is scattered across platforms that expose no API, and each source covers a different audience. An agent that reads only one is working from a partial picture.
- This server collects them into one structured response, so the agent sees the whole picture in a single call.
- Multi-source with graceful degradation: any one source failing does not break the pipeline. A source that cannot be read is reported as unavailable rather than filled in with guesses.
Quick start
Requires Python 3.8+. No pip install needed.
Claude Desktop / WorkBuddy (mcp.json)
{
"mcpServers": {
"hot-topics": {
"command": "python",
"args": ["/absolute/path/to/server.py"]
}
}
}
Then ask your agent: "What are today's trending topics in the AI工具 track?" or "Get reddit intel for the kids track and draft a post."
Example tool calls
{"name": "get_trending", "arguments": {"track": "AI工具"}}
{"name": "get_trending", "arguments": {"track": "EN_ai", "sources": ["google_trends"]}}
{"name": "get_reddit_intel", "arguments": {"track": "poa"}}
Hosted API (no install, works behind any firewall)
Don't want to run it yourself, or your network can't reach these sources? A hosted version runs 24/7 on our Singapore server — one HTTP GET, JSON back:
GET http://43.160.199.215/hotapi/v1/hot?track=EN_ai&key=YOUR_KEY
GET http://43.160.199.215/hotapi/v1/reddit?track=poa&key=YOUR_KEY
GET http://43.160.199.215/hotapi/v1/health
30-day access key: $9 → https://niebingyu.gumroad.com/l/hot-topics-api
Notes
- 10-minute in-memory cache per source; polite to upstreams.
- All sources are public endpoints used for research; respect each platform's terms.
- Protocol: MCP over stdio, JSON-RPC 2.0, protocol version 2024-11-05.
License
MIT