Skip to content

Restart and crash recovery

AIgency reconciles persisted run state against reality on every app restart, so completed work stays completed and only genuinely interrupted work gets rescheduled.

AgentDock Pro

Restart and crash recovery is part of AIgency, available only in AgentDock Pro.

Runs, tasks, and their history are durably persisted — see AIgency persistence — but the live provider processes behind a running task obviously don't survive an app restart or crash. Recovery is what reconciles the two.

What happens on restart

AgentDock doesn't try to resume a killed provider process. Instead, on startup it checks every run that was running or had tasks mid-execution against what's actually still alive:

  • A task that was genuinely completed before the restart stays completed — recovery never re-runs finished work.
  • A task that was interrupted mid-execution (no live process, no completion) is reset so the scheduler can pick it up again on the next pass, the same as any other eligible pending task.
  • Any open paused interval or in-progress wait state (waiting_capacity, waiting_interaction) is reconciled against current reality rather than left stale.

Why this matters

Without this reconciliation step, a crash mid-run could either silently lose track of finished work (re-running it unnecessarily) or leave a run stuck believing a task is still active when the process behind it is long gone. Recovery is what makes AIgency safe to leave running unattended.

Related