TradingAgents-Astock is a multi-agent investment research framework specialized for Chinese A-share markets.
The project addresses the limitation that the original TradingAgents framework was designed for US equities and cannot handle A-share-specific data sources, trading rules, and market dynamics. It solves this by implementing a deep specialization across three dimensions: data layer, agent roles, and trading constraints. The framework deploys seven AI analyst agents that conduct bull-and-bear debates to reach investment decisions. Data integration pulls from multiple free Chinese sources including mootdx, East Money, Sina, and Tonghuashun via direct HTTP connections. The analyst roles extend beyond the original four to include specialized agents for policy analysis, retail trader tracking, and share unlock monitoring. Trading logic incorporates A-share mechanics such as T+1 settlement, daily price limits, minimum lot sizes, and trading hours restrictions. The framework outputs Chinese-language investment reports while maintaining English-language internal reasoning to preserve analytical quality.
Developers should adopt this tool if they are researching multi-agent systems applied to Chinese equity markets or building investment analysis tools for A-share trading. It suits academic research, educational projects, and quantitative finance experimentation. The project explicitly disclaims that it provides no investment advice and constitutes no investment service. The framework is installable via pip with zero external service dependencies, making it accessible for local experimentation without cloud infrastructure.
Development activity shows consistent engagement with the specialized domain. The project maintains detailed documentation of modifications from the upstream TradingAgents framework. The codebase is structured around a dual-LLM design where quick-thinking models handle individual analyst roles while deep-thinking models manage portfolio-level decision synthesis. The implementation includes comprehensive handling of A-share-specific constraints and data formats throughout the agent pipeline.
