title	AgenticTrade
app_file	trading_floor.py
sdk	none

🤖 AgenticTrade – Autonomous AI-Powered Trading Floor

Traders today rely on rigid bots and static signals. AgenticTrade changes the game by launching autonomous AI agents, each inspired by legendary investors like Buffett, Soros, Dalio, and Wood. These agents research markets, generate trade ideas, and execute portfolio moves—fully automated and conversation-capable.

Built with Python, OpenAI LLMs, and the MCP (Model Context Protocol), AgenticTrade simulates a full trading desk with real-time decision-making and Pushover notifications.

---

🖼 Interface Preview

AI traders reviewing market conditions

Log tracer tracks all trading decisions

Live push notification to developer via Pushover API

---

🧪 Methodology

AgenticTrade is designed as a fully autonomous trading simulation platform, relying on tools and context agents. Here’s how it works:

1. Trader Initialization

   • Each trader is initialized with a name, model, and strategy.
   • They alternate between trade and rebalance mode.

2. Tool Use via MCP

   • Traders access tools like get_share_price, push, and research.
   • Tools run in separate MCP servers (market_server.py, push_server.py, etc.).

3. Market Awareness

   • market.py fetches data using the Polygon API.
   • Scheduler in trading_floor.py checks if the market is open.

4. Logging and Tracing

   • Custom tracer in tracers.py logs events using write_log.
   • Every trace has a unique ID tied to the trader.

5. Notifications

   • After trading, traders send a brief update to the developer via Pushover API.

---

📁 File Overview

Filename	Purpose
trading_floor.py	Runs all traders in a timed loop
traders.py	Trader logic using LLMs and MCP tools
market.py	Polygon-based share price fetcher
market_server.py	MCP server to respond with share prices
push_server.py	MCP server to send push notifications
reset.py	Resets traders to their original strategies
templates.py	Instruction templates per trader/agent
mcp_params.py	Tool configurations for MCP servers
tracers.py	Logs all trace and span activity

---

⚙️ Environment Variables

Create a .env file with the following keys:

POLYGON_API_KEY=your_polygon_api_key
POLYGON_PLAN=paid
PUSHOVER_USER=your_user_key
PUSHOVER_TOKEN=your_app_token
RUN_EVERY_N_MINUTES=60
RUN_EVEN_WHEN_MARKET_IS_CLOSED=false
USE_MANY_MODELS=true

---

🚀 How to Run

1. Install dependencies:

pip install -r requirements.txt

2. Start the autonomous trading floor:

python trading_floor.py

3. Reset strategies (optional):

python reset.py

---

👤 Trader Personas

Name	Role Model	Strategy Type
Warren	Warren Buffett	Long-term value investing
George	George Soros	Macro and contrarian bets
Ray	Ray Dalio	Risk parity + macro hedge
Cathie	Cathie Wood	Crypto + innovation focus

---

📦 Dependencies

openai
python-dotenv
requests
pydantic
asyncio
pypdf
firebase-admin
gradio (optional)

---

🔔 Real-Time Push Notifications

Each trader sends a push alert after finishing trades. Example:

💬 Warren bought 50 shares of BRK.B after identifying undervaluation. Portfolio remains stable with strong fundamentals.

---

📑 Summary

AgenticTrade combines LLM reasoning, market intelligence, and modular tools to simulate a real-world trading desk—autonomous, explainable, and intelligent. It’s the perfect platform to experiment with financial AI agents.

📄 Technical documentation and trading logs coming soon.