Nova AI On-Call can detect, investigate, and remediate incidents on its own. Autonomy levels control exactly how much it does without you. The safe path is to start in shadow mode, build trust, and graduate to autonomous remediation with clear guardrails.
Go to the Agent Autonomy page. This is where you set how far Nova can act for each class of action.
In shadow mode Nova watches, diagnoses, and suggests, but never acts. Compare its suggestions against what your team would have done to build confidence.
Next, let Nova propose actions that a human approves in the Approval Queue. You stay in control while Nova does the investigation work.
Set confidence thresholds so Nova acts on its own above a score and asks below it. High-confidence, low-risk fixes (such as restarting a crash-looping container) run automatically; ambiguous ones wait for you.
Arm the safety controls: blackout windows, the AI Cost Breaker, the Error Budget Gate, and approval for dangerous commands. These cap what autonomy can ever do.
Agent trust decays without positive evidence, and well-performing agents earn higher action tiers over time (4, then 20, then 50, then all). Watch the Agent Ledger to see trust move.
Turn on AI On-Call in shadow mode and watch Nova work before it acts.
Get Started Free