Krynos is an autonomous AI trading agent built for BTC/USD on Kraken. It combines a multi-agent debate workflow with quantitative market scoring to decide when to buy, sell, or hold. Two AI agents argue the bull and bear case, then a judge weighs their reasoning against live indicators such as EMA crossovers, RSI, MACD, Bollinger Bands, sentiment feeds, and PRISM market signals. The system can run in paper mode for safe testing or switch to live execution, while enforcing risk controls like trade-size caps, confidence thresholds, stop loss, take profit, and a daily circuit breaker. Every round is logged to SQLite and displayed in a Streamlit dashboard with trade history, debate logs, portfolio simulation, performance metrics, cross-asset intelligence, and live crypto news. The result is a transparent trading copilot that does more than place orders: it explains why each action was taken, shows how signals and AI reasoning interact, and gives users a clear view of strategy behavior in real time.
Category tags:"Demo link is not working. The 3 agent concept i.e Bull, Judge, Bear makes sense. In case of tie where priority will be given is unclear. "
Vasu Raj Jain
Senior Software Engineer
"Application of Technology 4/5 Bull/Bear debate + impartial judge LLM, quantitative signal scoring (±100 range), Groq Llama 3.3-70B, Kraken CLI, PRISM API, SQLite WAL mode, Streamlit dashboard. Clear and well-structured system diagram. Business Value 4/5 BTC/USD focus is narrow but appropriate for a hackathon demo. The transparent "explains why each action was taken" angle is good UX. Debate logs + trade history in dashboard shows accountability. Presentation 4/5 Clean README, ASCII system diagram, well-documented params table. PDF 12 slides (image-based, couldn't verify text claims). Live dashboard exists but redirected. Originality 3.5/5 Multi-agent debate pattern (similar to Swiftward Gamma) is a strong approach. The judge + quantitative signal scoring hybrid is distinctive. But the overall structure (Streamlit + AI signal + Kraken) is a common hackathon pattern. Standout elements: ✅ Bull/Bear debate with impartial judge LLM ✅ Quantitative scoring (±100) combining EMA, RSI, MACD, Bollinger Bands ✅ Real sentiment data (Fear & Greed, Binance funding, open interest, order book) ✅ PRISM API cross-asset signals ✅ Clear risk params: 5% daily circuit breaker, 2% per-trade cap, 1.5% stop loss, 1.2% take profit, 0.45 min confidence ✅ Paper/live toggle (one config flag) ✅ SQLite WAL mode for concurrent writes ⚠️ BTC/USD only — narrow scope ⚠️ No on-chain component (unlike Swiftward, HedgeFlow, Forge8004) Total: ~4.0/5 — Very clean and honest execution. The debate architecture is the standout, and the quantitative signal scoring as a gate before the debate is smart. Not as technically ambitious as Swiftward, but well put together."
Sanem Avcil
"Overview Krynos is a well-structured autonomous AI trading agent focused on BTC/USD using a multi-agent debate system combined with quantitative indicators. It feels like a balanced and thoughtful hackathon project that prioritizes transparency and explainability. Overall, it gives a clear impression of a functional, user-friendly trading copilot rather than just another black-box bot. Pros The multi-agent debate workflow (bull vs bear agents with a judge) adds intelligent reasoning layered on top of standard indicators like EMA, RSI, MACD, and Bollinger Bands, plus sentiment and PRISM signals. Strong emphasis on risk management with trade-size caps, stop loss, take profit, confidence thresholds, and daily circuit breaker, plus paper trading mode for safety. Excellent transparency through full logging to SQLite and a comprehensive Streamlit dashboard showing debate logs, trade history, and performance metrics. Cons The description lacks specifics on the underlying AI models used, how the judge agent makes final decisions, or any backtesting/live performance data. Integration details with Kraken for live execution are minimal. While the debate system sounds promising, there is no mention of how conflicts or low-confidence debates are resolved in volatile market conditions."
Anton Kiselev
Lead Backend Developer