TradingAgents-Astock is an Apache-2.0 Python-based framework for A-share multi-agent investment research, adapted from TradingAgents, with 7 analyst roles and data sources from MootDX and Chinese providers. It includes a CLI and a Streamlit web UI, and releases note breaking changes in v0.3.0; latest push was 2026-07-26.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (2 so far).
What it is
TradingAgents-Astock is an Apache-2.0 licensed Python project forked from TradingAgents, tailored for A-share markets. It implements a multi-agent research framework with 7 analysts and supports bull/bear debates, risk assessment, and a structured decision workflow. The project emphasizes data source integration from MootDX and Chinese providers, and offers both CLI and Web UI interfaces.
How it works
The framework combines multiple agents (analysts, researchers, traders, risk debaters, managers) to generate reports and run debates. It exposes data vendors for A 股 data, an LLM-driven decision process, and a two-stage thinking model (quick_think_llm for analysts and deep_think_llm for managers). The web UI provides a 12-stage pipeline visualization and exports Markdown or PDF reports. It supports OpenAI-compatible gateways and several LLM providers via a configurable env and config dict.
Getting started
To install and run:
git clone https://github.com/simonlin1212/tradingagents-astock.git
cd tradingagents-astock
pip install -e .
− Optional: install Google Gemini support:
pip install -e ".[google]"
Configure LLMs via a .env file with keys such as MINIMAX_API_KEY, DEEPSEEK_API_KEY, ZHIPU_API_KEY, DASHSCOPE_API_KEY, OPENAI_API_KEY, ANTHROPIC_API_KEY, OPENAI_COMPATIBLE_API_KEY, or BACKEND_URL for compatible gateways.
Run analysis through the CLI or Web UI. Web UI can be started with:
tradingagents-web
or
streamlit run web/app.py
Open http://localhost:8501 in the browser.
Recent releases
Latest release is v0.3.0 (2026-07-26): 删除可执行价位能力 and 8 项修正 per A 股 rules; note it introduces destructive changes for downstream code relying on fields like entry_price, stop_loss, position_sizing, price_target. Prior release v0.2.21 added features for customizable TA lookback and per-month analysis, and v0.2.20 added an OpenAI-compatible provider with a configurable base_url. The changelog details various fixes and updates across v0.2.x and v0.3.0.
Traction
Stars: 2658
Behind the repo
This is a fork of TauricResearch/TradingAgents aimed at A 股, with full Apache 2.0 license and no external service dependency beyond LLM API keys. It integrates data sources via mootdx and several Chinese providers and includes a structured 7-analyst workflow.
Caveats
License: Apache-2.0 Created: 2026-05-13 Last push: 2026-07-26