Docs/AI Trading Agents/Safety Guard Layer

Safety Guard Layer

Production

Budget, rate, anomaly, and kill-switch controls create a hard execution perimeter around all agents.

Objective

Prevent unsafe autonomous behavior without removing useful automation.

Budget and Limit Gates

Budget controls cap notional exposure per agent, strategy, and time window.

Permits and account constraints are rechecked before final execution.

Rate and Anomaly Controls

Rate limits prevent action bursts during unstable conditions. Anomaly checks score unexpected behavior against baseline patterns.

High-severity anomalies route to hold-and-review or hard-stop behavior depending on policy class.

Kill Switch

Kill switch immediately halts new execution attempts and can trigger cleanup workflows for open strategies.

The switch must be externally triggerable and observable in operator dashboards.

Production Checkpoints

  • Budget exhaustion blocks additional execution
  • Rate bursts are throttled with explicit reason codes
  • Anomaly-triggered holds are auditable
  • Kill switch state change propagates immediately

Continue exploring

AI Trading Agents