金融交易人工智能代理MCP
](https://badge.fury.io/py/finance-trading-ai-agents-mcp)  
免费、专业、开源的金融分析和定量交易MCP服务器。它允许您使用模拟真实金融公司运营的部门架构，通过一个命令部署本地金融MCP服务。它支持传统指标、价格行为分析、经济日历、基本面和新闻整合，提供与LLM和算法交易的无缝交互。
💬 社区与支持
 
加入我们的社区进行讨论、支持和更新！
✨ 特性
- 🚀 一键部署：快速启动本地金融MCP服务
- 🏢 部门架构：模拟真实世界的金融公司部门
- 📊 综合分析：传统技术指标+价格走势分析
- 📅 实时数据：经济日历、基本面、经纪商/证券信息和新闻整合
- 🤖 AI集成：针对LLM优化的接口
- ⚡ 高性能：实时流式OHLC数据处理
- 🔧 可扩展：支持自定义MCP服务
- MCP可以呼叫经纪商查询账户信息、下订单和取消订单。
- 🪩 支持本地RPC/PubSub服务。您可以跨进程/软件/代码语言通信。
最快体验
文档
- 完整文档：https://docs.aitrados.com/en/docs/finance-trading-ai-agents-mcp/quickstart/
财务数据的免费secret_key（AITRADO_secret_key）
在这里获取：https://www.aitrados.com/
基本呼叫工具提示词嵌入您的代理系统提示中
de.md 德语
en.md 英语
fr.md 法语
ja.md 日本语
kr.md 韩语
e.md 西班牙语
ru.md 俄语
zh_cn.md 简体中文
zh_tw.md 繁体中文
示例
示例/添加_自定义_mcp_示例
示例/mcp-clients
examples/run_mcp_examples
examples/env.example.py
📦 安装
来自PyPI（推荐）
pip install finance-trading-ai-agents-mcp来源
git clone https://github.com/aitrados/finance-trading-ai-agents-mcp.git
cd finance-trading-ai-agents-mcp
pip install -r requirements.txt
#pip install -e .🚀 快速开始
基本用途
from finance_trading_ai_agents_mcp import mcp_run
if __name__ == "__main__":
#from examples.env_example import get_example_env
#get_example_env()
mcp_run()或简单命令
# auto finding .env file and config.toml
finance-trading-ai-agents-mcp #or python -m finance-trading-ai-agents-mcp
#Specify .env file path
finance-trading-ai-agents-mcp --env-file .env中间使用
自定义MCP服务器和自定义MCP功能
from finance_trading_ai_agents_mcp import mcp_run
from examples.env_example import get_example_env
if __name__ == "__main__":
get_example_env()
from examples.addition_custom_mcp_examples.addition_custom_mcp_example import
AdditionCustomMcpExample
AdditionCustomMcpExample()
mcp_run()
# or
# mcp_run(addition_custom_mcp_py_file="./addition_custom_mcp_example.py")高级用法
与MCP服务器的实时WebSocket数据集成
from aitrados_api.universal_interface.callback_manage import CallbackManage
from finance_trading_ai_agents_mcp import mcp_run
from examples.env_example import get_example_env
"""
Real-time WebSocket Data Integration with MCP Server
This script demonstrates how to run an MCP (Model Context Protocol) server while simultaneously
receiving real-time WebSocket data streams, achieving data reusability across your project.
Key Benefits:
🔄 **Data Reusability**:
- The same real-time data feed serves both MCP clients (like Claude Desktop) and your custom application logic
- Eliminates duplicate API calls and reduces bandwidth usage
- Centralizes data management in one location
⚡ **Real-time Integration**:
- Multi-timeframe OHLC data: Real-time price feeds for trading analysis
- Event data: Market events, earnings, splits, etc.
- News data: Financial news updates as they happen
- Authentication: Connection status and auth events
- General messages: System notifications and other data
🏗️ **Architecture Advantages**:
- MCP server handles AI/LLM requests with structured financial data
- Custom callbacks process the same data for your trading algorithms
- WebSocket connection is shared, ensuring data consistency
- Thread-safe data management with proper synchronization
📊 **Use Cases**:
- AI-powered trading assistants with real-time market data
- Automated trading systems with LLM decision support
- Real-time portfolio monitoring with AI analysis
- Market research tools combining AI insights with live data
- Risk management systems with instant alert capabilities
🎯 **Practical Example**:
When Claude Desktop requests "Show me AAPL's current price", the MCP server provides real-time data.
Simultaneously, your custom callback can execute trading logic based on the same price update.
This eliminates the need for separate data feeds and ensures perfect synchronization.
⚙️ **Callback System**:
Each callback type handles specific data streams:
- multi_timeframe_callback: OHLC candlestick data for technical analysis
- event_handle_callback: Corporate actions and market events
- news_handle_callback: Breaking financial news
- auth_handle_callback: Connection and authentication status
- general_handle_callback: System messages and notifications
- show_subscribe_handle_callback: Subscription management events
This dual-purpose architecture maximizes the value of your real-time data subscription while
providing both AI capabilities and custom application logic in a single, efficient system.
"""
def multi_timeframe_callback(*args, **kwargs):
print("Multi-timeframe data received:", args, kwargs)
def event_handle_callback(client, *args, **kwargs):
print("Event data received:", args, kwargs)
def news_handle_callback(client, *args, **kwargs):
print("News data received:", args, kwargs)
def auth_handle_callback(client, *args, **kwargs):
print("Auth message received:", args, kwargs)
def general_handle_callback(client, *args, **kwargs):
print("General message received:", args, kwargs)
def show_subscribe_handle_callback(client, *args, **kwargs):
print("Subscribe handle message received:", args, kwargs)
def ohlc_chart_flow_streaming_callback(*args, **kwargs):
print("OHLC chart flow streaming data received:", args, kwargs)
def ohlc_handle_callback(client, *args, **kwargs):
print("OHLC handle message received:", args, kwargs)
def error_handle_callback(client, *args, **kwargs):
print("Error handle message received:", args, kwargs)
if __name__ == "__main__":
get_example_env()
# Register all custom callbacks
CallbackManage.add_custom_multi_timeframe_callback(multi_timeframe_callback)
CallbackManage.add_custom_event_handle_msg(event_handle_callback)
CallbackManage.add_custom_news_handle_msg(news_handle_callback)
CallbackManage.add_custom_auth_handle_msg(auth_handle_callback)
CallbackManage.add_custom_handle_msg(general_handle_callback)
CallbackManage.add_custom_show_subscribe_handle_msg(show_subscribe_handle_callback)
CallbackManage.add_custom_ohlc_chart_flow_streaming_callback(ohlc_chart_flow_streaming_callback)
CallbackManage.add_custom_ohlc_handle_msg(ohlc_handle_callback)
CallbackManage.add_custom_error_msgs(error_handle_callback)
mcp_run()命令行界面（CLI）
# auto finding .env file
python -m finance-trading-ai-agents-mcp #or finance-trading-ai-agents-mcp
#Specify .env file path
finance-trading-ai-agents-mcp --env-file .env
# --env-config
--env-config '{"DEBUG":"1","AITRADOS_SECRET_KEY":"YOUR_SECRET_KEY","OHLC_LIMIT_FOR_LLM":"20","RENAME_COLUMN_NAME_MAPPING_FOR_LLM":"interval:timeframe,","OHLC_COLUMN_NAMES_FOR_LLM":"timeframe,close_datetime,open,high,low,close,volume","LIVE_STREAMING_OHLC_LIMIT":"150","ENABLE_RPC_PUBSUB_SERVICE":"1"}'
# Show help
finance-trading-ai-agents-mcp --help
# Start the service
finance-trading-ai-agents-mcp --env-config '{"DEBUG":"1","AITRADOS_SECRET_KEY":"YOUR_SECRET_KEY"}'
# Run with custom MCP server and custom MCP functions from Python file
python -m finance_trading_ai_agents_mcp -c examples/addition_custom_mcp_examples/addition_custom_mcp_example.py --env-config '{"DEBUG":"1","AITRADOS_SECRET_KEY":"YOUR_SECRET_KEY"}'
# Specify port
python -m finance_trading_ai_agents_mcp -p 9000 --env-config '{"DEBUG":"1","AITRADOS_SECRET_KEY":"YOUR_SECRET_KEY"}'API使用示例
部门服务架构
finance_trading_ai_agents_mcp/
├── api/ # API interface layer
├── mcp_services/ # Core MCP services
│ ├── traditional_indicator_service.py
│ ├── news_service.py
│ └── global_instance.py
├── live_streaming_ohlc_operations/ # Real-time data streaming
├── parameter_validator/ # Parameter validation
├── mcp_result_control/ # Result control
├── addition_custom_mcp/ # Custom extensions
└── examples/ # Usage examples主要服务模块
- 传统指标服务：传统技术指标的计算
- 新闻服务：获取和分析财经新闻
- 直播操作：实时OHLC数据流处理
- API实例：RESTful API
- MCP经理：服务管理和协调
📊 功能模块
技术指标分析
- 移动平均线（SMA、EMA、WMA）
- 相对强弱指数（RSI）
- 布林线
- 指数平滑异同移动平均线
- 随机振荡器
- 更传统的指标。..
价格行为分析
- 支撑和阻力识别
- 图表模式识别
- 趋势分析
- 产量分析
基本面
- 财务数据采集
- 经济指标分析
- 公司基本面评估
新闻与情绪分析
- 实时财经新闻
- 情感分析
- 事件驱动分析
🛠️ 配置
环境变量
创建一个 .env 文件：
##debug
DEBUG=true
##Free Register at AiTrados website https://www.aitrados.com/ to get your API secret key (Free).
AITRADOS_SECRET_KEY=YOUR_SECRET_KEY
##Enable RPC/PubSub Service.you can cross process/software/code language communication.easily call api /websocks/other service
## see https://docs.aitrados.com/en/docs/api/trade_middleware/overview/
ENABLE_RPC_PUBSUB_SERVICE=0
##LIVE_STREAMING_OHLC_LIMIT:Real-time OHLC data stream length,default 150
##Prevent the strategy result from not being obtained due to insufficient ohlc length. For example, the value of MA200 can only be calculated when the length of ohlc is greater than 200.
LIVE_STREAMING_OHLC_LIMIT=149
#MCP LLM Setting
##OHLC_LIMIT_FOR_LLM :Due to the window context size limitations of the Large Language Model (LLM), please set a reasonable number of OHLC rows. This setting will only affect the output to the LLM and will not influence strategy calculations
##If it is a multi-period chart analysis, the OHLC_LIMIT_FOR_LLM adjustment is smaller
OHLC_LIMIT_FOR_LLM=30
##You can modify the ohlc column names to suit your trading system. Mapping example:name1:myname1,name2:myname2
RENAME_COLUMN_NAME_MAPPING_FOR_LLM=interval:timeframe,
##OHLC_COLUMN_NAMES_FOR_LLM:Filter out redundant column names for LLM input. The column names should be separated by commas.
OHLC_COLUMN_NAMES_FOR_LLM=timeframe,close_datetime,open,high,low,close,volume
📚 文档和示例
API文件
启动服务后，请访问： http://127.0.0.1:11999/
示例
请参阅 examples/ 更多示例的目录：
examples/basic_usage.py-基本用法examples/advanced_analysis.py-高级分析examples/custom_indicators.py-自定义指示器
🤝 贡献
我们欢迎各种形式的贡献！
- 分叉回购
- 创建要素分支(
git checkout -b feature/AmazingFeature) - 提交您的更改(
git commit -m 'Add some AmazingFeature') - 推到分支(
git push origin feature/AmazingFeature) - 打开拉取请求
开发设置
git clone https://github.com/aitrados/finance-trading-ai-agents-mcp.git
cd finance-trading-ai-agents-mcp
pip install -e ".[dev]"📄 许可证
此项目根据MIT许可证获得许可-请参阅 许可证 文件以获取详细信息。
🔗 链接
- github：https://github.com/aitrados/finance-trading-ai-agents-mcp
- PyPI：https://pypi.org/project/finance-trading-ai-agents-mcp/
- Wiki文档：https://github.com/aitrados/finance-trading-ai-agents-mcp/wiki
- 问题跟踪器：https://github.com/aitrados/finance-trading-ai-agents-mcp/issues
📞 支持
如果您遇到任何问题或有任何建议，请联系我们：
- 电子邮件：support@aitrados.com
- 问题：https://github.com/aitrados/finance-trading-ai-agents-mcp/issues
🙏 致谢
感谢所有为此项目做出贡献的开发人员和用户。
______________________________________________________________________
⭐ 如果这个项目对你有帮助，请考虑给它一颗星！