What it does TradingAgents-CN is a Chinese-localized fork of the Tauric Research multi-agent LLM trading framework. It wraps LLM-orchestrated stock analysis in a full web stack: FastAPI backend, Vue 3 frontend, MongoDB + Redis data layer, and Docker deployment. The project explicitly positions itself as a learning and research platform—no live trading, just synthetic experiments and report generation.

The interesting bit The README spends almost as much space on copyright warnings and licensing as on features. The project uses a hybrid license: core analysis code is Apache 2.0, but the app/ (FastAPI) and frontend/ (Vue) directories are proprietary and require commercial authorization. The authors are also publicly feuding with a third-party site (tradingagents-ai.com) they accuse of stealing their code. Drama aside, the technical pivot from Streamlit to a proper SPA/API architecture is the real structural change.

Key highlights

• Full A-share/China-market data integration via Tushare, AkShare, and BaoStock, with multi-level fallback chains for real-time quotes
• Multi-LLM provider support including OpenAI, Google AI, and domestic models with persistent model selection
• Professional report export in Markdown, Word, and PDF
• Docker multi-arch builds (amd64 + arm64) with GitHub Actions automation
• Simulated trading environment for strategy backtesting without capital risk

Caveats

• The backend and frontend are not open source under Apache; commercial use requires contacting the author for a separate license
• v2.0 is already in closed beta and will not be open-sourced due to “piracy issues”
• Data must be fully synced before analysis or results will be wrong; this is a documented footgun

Verdict Worth a look if you want a Chinese-market-aware LLM trading research sandbox and can tolerate the licensing split. Skip it if you need a fully open, production-ready trading system—or if legal ambiguity gives you hives.