TradingAgents-Astock is a multi-agent investment research framework specialized for China’s A-share market. It adapts the well-known TradingAgents architecture, in which AI analyst agents debate bull and bear cases before making a decision, and rebuilds the data layer, analyst roles and trading rules around how A-shares actually work.
The project is open source on GitHub with about 3,750 stars and active development in 2026. Instead of Yahoo Finance data and US market assumptions, it pulls free data directly from Chinese sources such as mootdx, Eastmoney, Sina and Tonghuashun. Its seven analyst agents add A-share-specific roles covering policy analysis, hot-money (short-term speculative capital) tracking and share-unlock monitoring, and the framework enforces local rules including T+1 settlement, daily price limits and minimum lot sizes. Internal debate runs in English for reasoning quality, while final reports are produced for A-share investors. The authors stress it is a research and teaching implementation, not investment advice. A web UI is included for running analyses interactively, and the modular LangGraph design makes it straightforward to add analysts, swap data sources or plug in different LLMs, from Claude to domestic Chinese models, depending on budget and availability.
TradingAgents-Astock is free and open source under the Apache-2.0 license. You only pay for any LLM API usage your configuration consumes.
Developers and finance researchers interested in multi-agent LLM systems who want to experiment with AI-driven research workflows on the Chinese A-share market.