siva-chidambaram12/kalshi-trading-bot

Kalshi AI Trading Bot is an automated trading system that places bets on Kalshi, a regulated prediction market platform where users trade on the outcomes of real-world events like elections, economic data releases, and sports results. The bot is described as educational and research-oriented. The system's distinguishing feature is its multi-agent architecture. Rather than a single AI model making decisions, it runs five specialized AI agents in parallel: a forecaster that assesses probability, a news analyst that scores incoming headlines for relevance, a bull researcher arguing for a trade, a bear researcher arguing against it, and a risk manager with veto power. The agents debate and the system aggregates their outputs into a consensus probability, penalizing confidence when the agents strongly disagree. Position sizing uses the Kelly Criterion, a mathematically-derived method for betting a fraction proportional to your estimated edge, with a default cap at 25 percent of the theoretical optimal to reduce variance. The bot includes a hard daily AI spending cap, paper trading mode for testing without real money, WebSocket streaming for real-time market prices, RSS news aggregation from major sources, and a category scoring system that tracks which types of markets the bot has historically performed well in. A developer interested in prediction market trading systems or AI-driven decision making research would use this. The tech stack is TypeScript on Node.js 22.5, using OpenRouter to access multiple language models and SQLite for trade logging.