An AI Research Department That Fights Over A-Shares

What it does TradingAgents-AShare is a web application that mimics a buy-side research department for Chinese A-share stocks. You type a natural-language request like “research Moutai short-term,” and the system dispatches specialized agents—covering fundamentals, sentiment, technicals, macro, and capital flows—to produce initial theses. Those feed into bull and bear researchers who argue in structured rounds, followed by risk-management review, ultimately yielding a report card with direction, confidence, target price, and stop-loss. The whole pipeline is visible in a web UI that tracks portfolios, schedules recurring analyses, and archives historical reports.

The interesting bit The unusual part is the debate drawer: a vertical, token-level streaming timeline where you watch bullish, bearish, and risk-control agents argue in real time, complete with round-by-round grouping and a highlighted verdict card. It turns the usually opaque LLM reasoning into a spectator sport of institutional role-play.

Key highlights

• Natural-language intent parsing: typing a request like “调研茅台短线” automatically resolves the symbol and investment horizon without form fields.
• Real-time bull/bear/risk debate visualization with token-level streaming and round-based timelines.
• Portfolio import, tracking dashboard, and scheduled batch analysis that auto-disables after consecutive failures.
• Frontend model switching across OpenAI, Anthropic, Gemini, DeepSeek, Moonshot, Zhipu, and SiliconFlow, with built-in warmup and connectivity tests.
• REST API and an OpenClaw skill for integration into external scripts or assistants.

Caveats

• The README repeatedly stresses the system is for academic research and technical demonstration only; it does not provide actual investment advice.
• New modules and the frontend are released under the PolyForm Noncommercial 1.0.0 license, so commercial use is restricted.
• Market data may carry delays or deviations, and the upstream core logic is inherited from TauricResearch/TradingAgents.

Verdict Worth a look if you are building multi-agent LLM orchestration tools, financial dashboards, or institutional workflow simulations. Skip it if you need battle-tested trading alpha or a permissive commercial open-source license.