Drawdown Guard is an autonomous AI trading agent designed to turn a client’s downside tolerance into an enforceable portfolio constraint. Investors can tell their portfolio manager, “I can afford to lose 10%.” But once the portfolio is built, that promise is usually just a number in a document. Markets move, positions change, options expire, and the portfolio can quietly drift beyond the limit. Drawdown Guard turns that promise into an active control. Every trading session, the agent reconciles the real portfolio, stress-tests its positions across a range of market declines, and calculates whether the client’s downside budget is still respected. When the portfolio becomes under-protected, it evaluates the live options market and adds the protection required to close the gap. The key is that the LLM can’t override the risk constraints. Deterministic code calculates the loss budget, stress scenarios, hedge quantities, eligible strikes, and hard trading limits. The LLM operates inside those boundaries, choosing between eligible protective puts and collars and explaining its reasoning. A final deterministic Gate must approve every order before it reaches the broker. Drawdown Guard is designed to adapt to portfolio changes. A new position triggers a hedge. A client sale makes an existing hedge redundant. An aging or expired option can reopen the risk gap. Every decision, order, fill, and AI recommendation is recorded in an auditable journal. The result is an autonomous risk layer that keeps a changing portfolio aligned with a fixed client promise, continuously and without human intervention.
Category tags: