AI-powered swap trading monitoring system using CrewAI multi-agent framework.
๐ Live Demo: prism.lisekarimi.com
PRISM monitors USD SOFR swap trading positions in real-time and automatically generates trading signals when profit/loss thresholds are hit.
Note: This demo is specifically configured for USD SOFR (Secured Overnight Financing Rate) swaps only.
5 AI Agents working together:
- Market Data Agent - Fetches current USD SOFR swap rates (2Y, 5Y, 10Y, 30Y)
- Position Manager Agent - Loads trader's swap positions from database
- Risk Manager Agent - Analyzes market conditions and sets dynamic profit/loss thresholds
- Risk Calculator Agent - Calculates P&L for each position
- Trading Decision Agent - Generates CLOSE/HOLD signals based on thresholds
Thresholds:
- Close position when profit โฅ $50K
- Close position when loss โค -$25K
- Otherwise HOLD
- CrewAI - Multi-agent orchestration
- PostgreSQL - Position & rate storage (Neon)
- Gradio - Web dashboard
- Python 3.11 - Runtime
Development Tools:
- Python 3.10+
- uv package manager
- Docker Desktop
- WSL (Windows Subsystem for Linux)
- Make:
winget install GnuWin32.Make(Windows) |brew install make(macOS) |sudo apt install make(Linux) - PostgreSQL database (you can use https://neon.com/ which is free)
- OpenAI API key for LLM
- Serper API key from https://serper.dev/ for Google Search API
-
Clone the Repository:
git clone https://github.com/lisekarimi/prism.git cd prism -
Set Up Environment:
- Copy the
.env.exampleto.envand fill in the required environment variables.
- Copy the
make devswap_positions- Trader's swap portfoliomarket_rates- Historical rate datatrade_signals- AI-generated trading decisions

