Forking a trading agent framework, then forbidding it to trade

What it does

TradingAgents-CN is a Chinese-market adaptation of Tauric Research’s multi-agent LLM stock analysis framework. It swaps the original Streamlit interface for a FastAPI backend and Vue 3 frontend, wires in domestic data sources like AKShare and Tushare for A-shares, and wraps the whole thing in Docker containers. The platform simulates trading strategies and exports research reports, but the authors repeatedly insist it is for education and research only, not for live trading.

The interesting bit

The project wears its legal caution on its sleeve: it ships with a mixed license that keeps the FastAPI backend and Vue frontend proprietary while open-sourcing the rest under Apache 2.0, and the maintainers have already declared that the upcoming v2.0 will stay closed-source because third parties allegedly pirated their code. That makes it less a traditional open-source community project and more an open-core product with a WeChat support channel.

Key highlights

• Replaces the upstream Streamlit UI with a FastAPI + Vue 3 stack, adding user auth, role management, and SSE/WebSocket progress tracking.
• Integrates Chinese market data via AKShare, Tushare, and BaoStock, with multi-level fallback chains for real-time quotes.
• Supports multiple LLM providers—including OpenAI, Google AI, and domestic endpoints—with a persistent model selector and an aggregated AiHubMix channel.
• Includes a mock trading system, multi-format report export (Markdown/Word/PDF), and a stock screener for strategy experimentation.
• Maintainers manually cherry-pick upstream updates rather than merging automatically, citing selective absorption.

Caveats

• The backend (app/) and frontend (frontend/) directories are not open source under Apache 2.0; commercial use requires a separate license, and v2.0 will not be publicly released at all.
• Documentation and guides are heavily tied to the WeChat ecosystem (WeChat articles, Bilibili videos), which may be inaccessible or inconvenient for non-Chinese-speaking developers.
• The README warns that stock data must be fully synchronized before analysis; otherwise results will contain data errors.

Verdict

Worth a look if you are a Mandarin-speaking developer or quant researcher who wants a containerized sandbox to experiment with multi-agent LLM analysis on Chinese equities. Skip it if you need a fully open-source stack, live trading execution, or an English-first developer experience.