How the Auto-Pilot agents actually decide to trade
Published July 13, 2026 · Updated October 8, 2026
There's no crystal ball inside a Tradexa agent — just a disciplined pipeline that runs the same way every time, without fear or greed. Here's the whole thing.
1. Scan
On a schedule, the agent pulls a fresh universe of candidates — momentum leaders, oversold dip setups, long-term trend names, and (if enabled) options. It's looking across the market so you don't have to.
2. Score
Each candidate gets a score from a stack of objective factors: trend (above/below the 50- and 200-day), moving-average alignment, RSI, MACD crossovers, 20-day breakouts, one-month trend, relative strength vs the S&P, and whether it's overextended. A news read can nudge it up or down. Only names that clear your strategy's threshold move on.
3. Enter
When a setup qualifies, the agent checks its budget and position limits. A real-money buy becomes a request for you to review and approve or skip. Opening a notification does not place a trade. Practice uses simulated orders. The trade log records what happened.
4. Manage the exit
For supported bot-owned real-money positions, the agent checks the configured stop-loss and take-profit rules and can send an exit automatically. That depends on the engine, market data and broker being available; a configured level is not a guaranteed fill price. Positions bought outside Tradexa remain view-only here. Review the exact settings on your agent rather than assuming every mode uses the same exit rules.
The honest part: a great process still takes losing trades. Scoring high is not a prediction — it's a filter. The edge, if any, comes from doing the same disciplined thing thousands of times, not from being right on any single trade.