PrimoAgent is a multi-agent AI system for comprehensive daily stock analysis and next-day price predictions. Built on LangGraph, it targets traders and researchers seeking to enhance decision-making by integrating Natural Language Processing (NLP), technical analysis, and portfolio management, aiming to minimize risk and potential losses through deep financial market analysis.

How It Works

The system employs a sequential LangGraph pipeline orchestrating four specialized agents. A Data Collection Agent gathers market data and news via APIs. A Technical Analysis Agent computes indicators like SMA, RSI, and MACD. A News Intelligence Agent quantifies NLP features from financial news. Finally, a Portfolio Manager Agent synthesizes these analyses into BUY/SELL/HOLD signals with confidence levels, using historical context for adaptive strategies. This modular, agent-based approach allows for deep financial market analysis.

Quick Start & Requirements

Setup involves creating and activating a virtual environment, then installing dependencies via pip install -r requirements.txt. Configuration requires populating a .env file with API keys for OpenAI, Finnhub, Firecrawl, and Perplexity. The workflow is a two-step process: first, run python main.py to collect and analyze data, interactively prompting for stock symbols and date ranges; second, run python backtest.py for backtesting.

Highlighted Details

Backtesting results claim varying effectiveness against Buy & Hold strategies. For META, PrimoAgent reportedly achieved a 31.97% return (Sharpe 2.899, Max DD 8.99%), outperforming Buy & Hold (22.16% return, Sharpe 1.124, Max DD 34.04%). NFLX also showed strong PrimoAgent performance (28.61% return, Sharpe 2.570). Conversely, AAPL and TSLA exhibited negative returns with PrimoAgent, though with substantially lower volatility and Max Drawdown compared to Buy & Hold.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or roadmaps are provided in the README.

Licensing & Compatibility

The project is licensed under the MIT License. It is explicitly stated as academic research code intended for educational purposes only and does not constitute financial advice or trading recommendations. Compatibility for commercial use is not explicitly addressed beyond the MIT license terms.

Limitations & Caveats

The trading strategies and analyses are experimental and have not been validated for real-world trading. Users must be aware that trading involves substantial risk of loss, and professional financial consultation is advised before making investment decisions.