Skip to main content
Back

TrendRadar Hits 61K Stars Aggregating Chinese and Global News With AI

A 61k-star open source tool aggregates hot news from 11 platforms, filters with AI, and pushes to 9 messaging channels.

Repo, Full Access
  • TrendRadar hit 61k stars as an open source AI news aggregator covering 11 Chinese platforms plus RSS.
  • Nine push channels supported: WeChat, Feishu, DingTalk, Telegram, email, ntfy, Bark, Slack, generic webhooks.
  • AI filtering uses natural language interest descriptions instead of keyword rules, with score thresholds.
  • LiteLLM integration supports 100+ LLM providers with automatic retries and fallback models.
  • MCP server exposes 17 tools for Claude Desktop, Cursor, and Cherry Studio to query history via natural language.
  • Three push modes (daily, current, incremental) and five scheduling presets fit different monitoring workflows.

Anyone who has tried to keep up with news across Weibo, Zhihu, Bilibili, Wall Street Journal knockoffs, and RSS feeds knows the problem: you spend more time doomscrolling than reading anything useful. TrendRadar is an open source attempt to fix that with an AI-driven aggregator that pulls hot topics from multiple platforms, filters them by your interests, and pushes a summary to whatever messaging app you already check.

The project has climbed past 61,000 stars on GitHub and ships as a Python service you can run via Docker, GitHub Actions, or locally with uv. It supports nine push channels including WeChat, Feishu, DingTalk, Telegram, email, ntfy, Bark, Slack, and generic webhooks, plus an MCP server so you can query the data with natural language from Claude Desktop or Cursor.

What it actually does

By default it monitors 11 mainstream platforms including Zhihu, Douyin, Bilibili hot search, Wall Street CN, Tieba, Baidu hot search, Cailianshe, The Paper, ifeng, Toutiao, and Weibo. Data comes from the newsnow API, and you can add or remove sources in config.yaml. RSS and Atom feeds get the same treatment, using shared keyword grouping so hot lists and subscriptions merge into one clean push.

The filtering syntax in frequency_words.txt is where the tool earns its keep. You can write plain keywords, mark required terms with