Durable state machines for agent work
3 Jul 2026 · 7 min read
How durable state machines make long-running agent work resumable, observable, and governable.
Durable state machines give agent workflows a source of truth that survives retries, deploys, tool failures, and human approval delays.
AstraRunner treats this as an execution-platform concern: the run needs durable state, traceable decisions, cost visibility, and enough structure for teams to review what happened after the agent finishes.
Patterns that work
- Represent waiting, blocked, failed, approved, and completed as distinct states.
- Store transition reason and actor for every state change.
- Make external side effects happen after checkpoint writes.
- Expose state history in the operator UI.
Implementation checklist
- Define legal transitions.
- Reject duplicate transition requests.
- Make every transition idempotent.
- Connect transitions to trace spans.
Watch outs
- Boolean status flags collapse important failure modes.
- Workers should not be the source of truth.
- Manual database edits destroy auditability.
Related reading
How this fits the Astra stack
AstraRunner owns the operational path for agent work: orchestration, scheduling, approval gates, traces, costs, and role-agent handoffs. AstraMemory owns durable context and retrieval. AstraGenie turns those capabilities into product workflows for teams that want automation without rebuilding the platform layer.