🧠 Autonomous Alpha Agents Trading Simulator

A Multi-Agent, Multi-MCP, LLM-Powered Stock Trading System

App link: https://huggingface.co/spaces/akankshakusf/autonomous-trading-sim

---

---

🚀 Project Overview

This project implements a fully autonomous multi-agent trading simulator leveraging LLMs (Large Language Models), the Model Context Protocol (MCP), and modular microservices.
It enables agents to research, decide, execute trades, and send notifications in a realistic, modular, and resilient simulation.

---

🛠️ Key Components

• Chat Interface (UI):
  User submits trading queries (e.g., "Buy 10 shares of TSLA").

• LLM Trader Agent:
  Interprets user intent, orchestrates the workflow, and delegates tasks to specialized MCP servers.

• MCP Servers:

  • Market MCP: Retrieves real-time or fallback share prices and market status.
  • Account MCP: Handles buying/selling, funds verification, and account state updates.
  • Push MCP: Sends push notifications upon trade success or failure.

• Resilience:
  Handles all failure modes (market closed, insufficient funds, price unavailable, push notification failure, or MCP server unavailable) and provides user-facing feedback for each scenario.

---

📊 Example Flow

1. User:
   "Buy 10 shares of TSLA"

2. LLM Trader Agent:

   • Checks if the query can be answered directly or if market/account/push MCP interaction is needed.
   • Queries the Market MCP for current price and market status.
   • Calls Account MCP to execute trade if possible.
   • Invokes Push MCP to notify user upon trade completion.
   • Handles and returns errors if any MCP server fails (with explicit error messaging).

3. User:

   • Sees a clear response:
     • “TSLA shares bought!” (Success)
     • “Trade failed: insufficient funds”
     • “Trade failed: market closed”
     • “Trade failed: price unavailable”
     • “Push notification failed”
     • “Trade failed: system unavailable”

---

📦 Tech Stack

• Python (Gradio, FastAPI, Pydantic, asyncio, sqlite)
• OpenAI/LLM/Agents (Orchestration via custom agentic framework)
• MCP (Model Context Protocol) – Modular tool/server-based integration
• Polygon.io – Market data (with fallback logic)
• Pushover – Push notification microservice

---

💼 How Companies Can Use This System

1. Automated Portfolio Management:
   Deploy autonomous agent teams to manage real or simulated investment portfolios, allowing for continuous, data-driven trading decisions across global markets.

2. Rapid Prototyping of Trading Strategies:
   Experiment with new trading logic, agent behaviors, or market data sources in a modular, safe environment before rolling out to production.

3. Resilient, Auditable Trade Execution:
   Leverage the built-in error handling and logging for compliance, real-time monitoring, and robust recovery from system or third-party failures.

4. Intelligent Decision Support:
   Integrate the platform into existing finance operations to provide recommendations or real-time “what-if” analysis using AI-powered agents.

5. Customizable Multi-Agent Simulations:
   Model and test different trader personas, risk policies, or notification flows for research, training, or investor education—simply by swapping agent logic or microservices.

---