Aegis AI

Problem:

Most “AI trading bots” are demos that skip what makes trading systems trustworthy in production: auditable decision making, strict risk controls, recoverable state, and clear operator visibility. Teams end up with bots that trade without explainability, lose context after restarts, and provide frontends with developer-only logs.

Solution:

Aegis AI is an end-to-end trading engine built around explainability and safety. It generates trade candidates using an LLM (Groq), runs every decision through a dedicated risk layer (position sizing, stop-loss/take-profit, and daily-loss circuit breaker), and executes through Kraken CLI in paper or live mode. Positions are persisted in MongoDB and recovered on boot so monitoring continues after restarts. A Socket.IO event system emits human-readable, structured updates so a non-technical user can understand exactly what the bot is doing and why.

Why it’s valuable:

Aegis AI bridges “AI reasoning” with real execution constraints:

- Operators get real-time, user-friendly explanations (not JSON dev logs).

- Risk gating is explicit and measurable (dailyPnL + circuit breaker).

- The system is resilient (DB recovery + watcher lifecycle).

- Optional on-chain integration allows trade intents to be simulated/approved and then attested for validation/reputation tracking.