Pre / In / Post-flight

Gavrun governs the full action lifecycle in three phases.

01 · Pre-flight

Can this agent run and return good ROI?

Checked before execution begins. Defined in the policy editor.

Tool and model allowlist
Budget and token limits
Approved time windows
Priority vs other agents (tenant-wide pool)

Outcomes: Allow / Defer / Block

02 · In-flight

Is this action safe and worth it?

Checked on every governed tool call.

Prompt and payload validation (injection detection)
PII and sensitive data scanning
Model fit and expected ROI check
Human approval requirements

Outcomes: Allow / Hold / Modify

03 · Post-flight

Can we explain and improve it?

Decision trail and explainability (audit log)
Spend and outcome monitoring (productivity page)
Privilege de-escalation when an agent misbehaves
Feedback loop back into pre-flight policy

Outcomes: Close the loop

The full lifecycle: pre-flight → in-flight → post-flight → smarter pre-flight