TradingAgents-CryptoA

Multi-Agent LLM Analysis Framework for A-Shares + Crypto

English | 中文

---

Overview

TradingAgents-CryptoA is an enhanced fork of TauricResearch/TradingAgents, adding support for Chinese A-share market and cryptocurrency multi-agent analysis.

The framework deploys specialized LLM-powered agents (fundamental analysts, sentiment experts, technical analysts, traders, risk management teams) that collaboratively evaluate market conditions and produce trading decisions.

Comparison: Original vs CN vs CryptoA

	Original	CN Version	CryptoA (this project)
A-Share	❌	✅ Tushare / AKShare / BaoStock	✅ AKShare (free, no registration)
Crypto	❌	❌	✅ CCXT (Binance / OKX / Bybit)
US Stocks	✅ yfinance / Alpha Vantage	✅	✅ inherited
On-Chain Agent	❌	❌	✅ whale tracking / open interest
Policy Agent	❌	❌	✅ PBOC / CSRC policy analysis
Fund Flow Agent	❌	❌	✅ northbound / institutional flows
CN Sentiment Agent	❌	✅	✅ Eastmoney / Xueqiu sentiment
News Filter	❌	✅ multi-layer + quality scoring	✅ dedup / ad removal / freshness / relevance
LLM Providers	OpenAI	OpenAI / Gemini / DeepSeek / Qwen	8 providers (OpenAI / Claude / Gemini / DeepSeek / Qwen / xAI / OpenRouter / Ollama)
WebUI	❌	✅ Vue 3 + Element Plus	❌ CLI-first
DB Cache	❌	✅ MongoDB + Redis	❌ file-based cache
Docker	❌	✅ multi-arch	❌ local execution
License	Apache 2.0	Mixed (WebUI requires commercial license)	Apache 2.0 (fully open source)

Strengths

• 🔗 Only TradingAgents fork with cryptocurrency support
• 🆓 Zero-cost data — AKShare + CCXT are free, no paid data subscriptions required
• ⛓ On-chain analysis — unique on-chain agent for DeFi / derivatives traders
• 🔌 Widest LLM compatibility — 8 providers + local Ollama
• 📜 Fully open source — Apache 2.0, no commercial license restrictions

Current Limitations

• 🖥 No WebUI yet (CLI-first; may be added later)
• 🐳 No Docker deployment yet (planned for a future release)
• 📊 A-share financial data relies on AKShare; fields may be less comprehensive than Tushare
• 🧪 Early stage — agent prompts and workflows are under active optimization

Quick Start

Installation

git clone https://github.com/OK-computer008/TradingAgents-CryptoA.git
cd TradingAgents-CryptoA

# Create virtual environment
conda create -n tradingagents python=3.13
conda activate tradingagents

# Install dependencies
pip install -r requirements.txt

Configure API Keys

Copy the environment template and fill in your API key(s):

cp .env.example .env

You only need to set the key for the LLM provider you plan to use.

Run A-Share Analysis

python run_a_share.py

Default target: 000060.SZ. Modify the ticker and date as needed:

from tradingagents.graph.trading_graph import TradingAgentsGraph
from tradingagents.default_config import get_a_share_config

config = get_a_share_config()
config["llm_provider"] = "deepseek"
config["deep_think_llm"] = "deepseek-chat"
config["quick_think_llm"] = "deepseek-chat"

ta = TradingAgentsGraph(debug=True, config=config)
_, decision = ta.propagate("600519.SH", "2026-03-20")
print(decision)

Run Crypto Analysis

python run_crypto.py

from tradingagents.graph.trading_graph import TradingAgentsGraph
from tradingagents.default_config import get_crypto_config

config = get_crypto_config(exchange="binance")
config["llm_provider"] = "anthropic"
config["deep_think_llm"] = "claude-sonnet-4-5"
config["quick_think_llm"] = "claude-haiku-4-5"

ta = TradingAgentsGraph(debug=True, config=config)
_, decision = ta.propagate("BTC/USDT", "2026-03-20")
print(decision)

Supported LLM Providers

Provider	Env Variable	Example Models
OpenAI	OPENAI_API_KEY	gpt-5.2, gpt-5-mini, gpt-4.1
Anthropic	ANTHROPIC_API_KEY	claude-sonnet-4-5, claude-haiku-4-5
Google	GOOGLE_API_KEY	gemini-2.5-pro, gemini-2.5-flash
DeepSeek	DEEPSEEK_API_KEY	deepseek-chat, deepseek-reasoner
DashScope (Qwen)	DASHSCOPE_API_KEY	qwen-max, qwen-plus, qwen-turbo
xAI	XAI_API_KEY	grok-4-0709
OpenRouter	OPENROUTER_API_KEY	Any OpenRouter-supported model
Ollama (local)	None required	Any locally deployed model

Analyst Combinations

A-Share Mode

market + fundamentals + cn_sentiment + policy + fund_flow + news

Crypto Mode

market + onchain + news

US Stock Mode (upstream default)

fundamentals + sentiment + news + technical

Project Structure

tradingagents/
  agents/          # Multi-agent definitions (analysts, researchers, trader, risk)
  dataflows/       # Data source adapters (AKShare, CCXT, yfinance)
  graph/           # LangGraph workflow orchestration
  llm_clients/     # Multi-LLM client factory
  default_config.py
run_a_share.py     # A-share analysis example
run_crypto.py      # Crypto analysis example

Acknowledgments

This project is built on TauricResearch/TradingAgents. We thank the original authors for their open-source contribution.

Original paper citation:

@misc{xiao2025tradingagentsmultiagentsllmfinancial,
      title={TradingAgents: Multi-Agents LLM Financial Trading Framework},
      author={Yijia Xiao and Edward Sun and Di Luo and Wei Wang},
      year={2025},
      eprint={2412.20138},
      archivePrefix={arXiv},
      primaryClass={q-fin.TR},
      url={https://arxiv.org/abs/2412.20138},
}

Disclaimer

This project is for educational and research purposes only. It does not constitute investment advice.

• Trading decisions are influenced by many factors including LLM model choice, data quality, and market conditions
• The project maintainers bear no responsibility for any losses incurred from live trading using this framework
• Please use this project with full awareness of the risks involved
• A-share and cryptocurrency markets carry significant risk — comply with local laws and regulations

License

Apache 2.0