How Cortiq Actually Works

The agent pulls D1 down to M15 candles from your MT5 history and runs a structured search for edges — train-only hypotheses, tested out-of-sample, written up in a report you can read. Nothing here touches live capital.

Findings become explicit, checkable rules: what the setup looks like, where you enter, how you size and manage risk, and what invalidates the trade. Preparation packages, sentiment reads and an instrument profile attach here, so a session starts from real context instead of a blank screen.

The playbook gets replayed against historical M1 data using the same AI decision cycle it would run live — strictly no-lookahead, with prep and sentiment fed in as-of the candle being tested, not with hindsight. This is where you find out whether the idea has any expectancy before a cent is at risk.

The same loop, promoted to real execution. The agent wakes on a candle close, a price level, or a scheduled time, reads the market, and opens, modifies or closes trades on your MT5 account — inside guardrails it cannot override.

Every session and backtest writes what it learned back into that instrument's knowledge ledger. Observations only become trusted once enough evidence backs them, live reads count for less than backtested ones, and a strategy's mastery rating rises only once it has been proven in production.