Crypto Trading Agents

中文文档

🙏 Acknowledgements

This project is based on TradingAgents by the Tauric Research team, as well as the paper arxiv.org/pdf/2412.20138. We extend our sincere thanks for their contributions!

In addition, the following authors and repositories also inspired this project: | Author | Repository | | --- | --- | | @delenzhang | TradingAgents | | @hsliuping | TradingAgents-CN |

✨ Key Features

💰 Crypto-Focused
Built upon the TradingAgents framework, specifically adjusted for cryptocurrency assets.
📈 Integrated Technical Analysis
Pulls data from professional technical analysis platforms instead of relying solely on LLM interpretation, reducing unreliable qualitative fluff. Supported Infomation Sources
📰 Targeted News Sources
Collects data from sources frequently used by crypto traders — reliable, relevant, and free! Supported Infomation Sources
❤️ Tailored to Your Trading Style
You can define custom investment preferences—whether you're an aggressive trader or a long-term investor, your style and strategy will be reflected in the report. Configure Investment Preferences
🚀 Incorporate External Reports
You can provide external researches or opinions for consideration—more context leads to better insights. Steps to Use
📄 PDF or Markdown Report Generation
Analysis reports are generated in readable formats.
📧 Scheduled Email Delivery
Combine with OS-level task schedulers to automatically generate and email reports - get market updates like you're the head of a trading desk.
🎥 Real-Time Report Logging
Generated reports are written to log files under ./logs in real time. Even if the process is interrupted or encounters an error, the partial report remains available—ensuring your API usage is never wasted.
⚙️ Fully Customizable
Easily modify or add new data sources. A detailed guide is provided to help you quickly adapt the tool to your needs. Customization

🛠️ Usage Guide

Installation

Clone the repository:

git clone https://github.com/Tomortec/CryptoTradingAgents.git
cd TradingAgents

Create a virtual environment:

conda create -n tradingagents python=3.13
conda activate tradingagents

Install dependencies:

pip install -r requirements.txt

Configuration

1. Configure LLM API Key

Create a .env file under the ./cli directory using .env.example and fill in your LLM API key, such as:
For Qwen: DASHSCOPE_API_KEY=XXXXXX
For ChatGPT: OPENAI_API_KEY=XXXXXX

See Supported LLMs and API Key Naming

2. Configure Information Source API Keys

Also add the required API keys for data sources into the ./cli/.env file

See Supported Information Sources

3. Check and Modify Configuration

Edit ./tradingagents/default_config.py to change the language, LLM settings, and other default configurations.

4. (Optional) Configure Investment Preferences

Create a file named investment_preferences in the ./cli directory to define custom investment preferences.

5. (Optional) Configure Email Sending

Set send_report_to_email = True in default_config.py,
then copy ./mailsender/.env.example to .env and fill in the email settings.

Running the Program

You can use CLI Mode or Script Mode.
CLI mode includes an interactive terminal interface; script mode is ideal for automation (e.g., hourly scheduled reports).

CLI Mode

Execute the main program from terminal:

python -m cli.main

Steps to Use

1. Enter Asset Symbol, such as BTC or ETH
2. Enter Analysis Date
3. Select Analyst Team - Market Analyst, Social Media Analyst, News Analyst and Fundamentals Analyst
4. Choose Research Depth
5. Import External Reports: Type y and press Enter to open the default editor, where you can input external viewpoints for the model to consider. Save the file when done.
6. Import Investment Preferences: Use the saved file at ./cli/investment_preferences or input them directly in the editor (optional).
7. Select LLM Model
8. Generate Report: After processing, the report will be saved under ./tradingagents/reports.

Script Mode

1. Edit ./cli/run.py as needed (e.g., set ticker or date)
2. Run the script:

python -m cli.run

Supported LLMs

| Name | API Variable | Tested | | ------------------- | ------------------- | ------ | | Qwen (by Alibaba) | DASHSCOPE_API_KEY | ✅ | | ChatGPT (by OpenAI) | OPENAI_API_KEY | ✅ |

Supported Information Sources

|Source|Name|API Variable|Data Type|Registration| |---|---|---|---| ---| | Alternative.me|Fear & Greed Index|None needed| Sentiment| N/A| | Binance | K-line, market depth, 24h price change, long/short ratio|None needed| Market| N/A| | Blockbeats| Blockbeats News| None needed| News| N/A| | CoinDesk| CoinDesk News| COINDESK_API_KEY| News| API Key Registration | | CoinStats| CoinStats News| COINSTATS_API_KEY| News|API Registration| | Reddit| Reddit Posts| REDDIT_CLIENT_ID, REDDIT_CLIENT_SECRET, REDDIT_USERNAME, REDDIT_PASSWORD, REDDIT_USER_AGENT | Sentiment & News | Register App| | taapi.io| Technical indicators like EMA, MACD, RSI, Supertrend, Bollinger Bands, Three White Soldiers, etc. | TAAPI_API_KEY| Technical Analysis | My Account |

Customization

Customize Prompts

Edit files under ./tradingagents/i18n/prompts

Customize Data Sources

Refer to ./tradingagents/dataflows/README.md

🔄 Planned Updates

• [x] Add LLM search capabilities for richer information retrieval
• [x] Enable automatic report delivery
• [ ] Integrate with freqtrade for backtesting/simulated trading
• [ ] Provide more LLMs, such as DeepSeek (use Qwen's Embedding)
• [ ] Improve prompt templates using latest LLM research
• [ ] Provide a UI interface
• [ ] ~~Integrate other price forecasting tools~~ (You can implement your own with Customize Data Sources. For forecasting tools, see CryptoMamba, Cryptopulse, etc.)

⚠️ Disclaimer

This project is for research and educational purposes only and does not constitute investment advice. Investing involves risk—make decisions cautiously.

<br/>

---

We welcome contributions! Including but not limited to submitting issues, fixing bugs, adding features, improving documentation, and localization.
⭐️⭐️ If this project helps you, please consider giving us a star! ⭐️⭐️