AXME CLOUD
Four waiting states that keep your agent alive across hours, days, and restarts
Agents hang, timeout, or lose state when waiting for slow humans or external tools. Standard async patterns don't persist across process restarts.
WAITING_FOR_HUMAN, WAITING_FOR_TOOL, WAITING_FOR_AGENT, WAITING_FOR_TIME — AXME's intent lifecycle holds state durably so your agents never lose work.
Why waiting state is a first-class concern
Production agents rarely finish in one shot. They wait for a manager to approve a quote, for Stripe to confirm a payment, for another agent to finish research, or until a scheduled window opens. If that wait lives only in process memory, a deploy or OOM kill loses the run — and you rebuild polling loops to compensate.
AXME Cloud models four explicit waiting states on every intent. The agent process can exit; the runtime holds durable state and wakes the workflow when the world is ready. Timeouts, reminders, and escalation paths attach to the wait itself — not to ad hoc cron jobs.
Four durable waiting states
CAPABILITIES
How it works.
WAITING_FOR_HUMAN
Approval, review, or clarification — agent sleeps until response.
WAITING_FOR_TOOL
External API or long-running tool — no busy loops.
WAITING_FOR_AGENT
Peer agent handoff via shared intent state.
WAITING_FOR_TIME
Scheduled resume after delay or cron.
DEEP DIVE
Production patterns.
Crash recovery
State survives deploys and OOM kills.
Timeout + escalation
Reminders and fallback assignees.
Cross-restart resume
Pick up exactly where you left off.
Designing waits in production
Use WAITING_FOR_HUMAN when a person must approve, review, or clarify — the intent stays in that state until a response or timeout. WAITING_FOR_TOOL covers long external APIs and batch jobs without busy loops in your agent. WAITING_FOR_AGENT hands work to a peer through shared intent state. WAITING_FOR_TIME schedules resume after a delay or cron boundary.
Pair waits with delivery modes: stream updates to a dashboard while waiting, push a webhook when the human acts, or poll from a mobile client — same intent, different bindings.
Common questions
- How long can an intent stay in a waiting state?
- Hours or days — state is persisted in AXME Cloud, not in your agent process. Configure timeouts and escalation per wait type.
- What happens if my agent crashes while waiting?
- The intent remains in the waiting state. A new worker resumes from the last durable checkpoint — no manual reconciliation.
- Can one intent have multiple waits in sequence?
- Yes. The lifecycle moves through processing → waiting → processing until completed or failed, with full audit on each transition.
Related reading
Deeper dives from the AXME blog.
Your Agent Has Been Stuck for 3 Hours and Nobody Knows
API call hangs. Agent blocks. No timeout fires. No one gets paged. The default behavior in most agent frameworks is silence. Here's how to fix that.
Read post →Why Your AI Agent Shouldn't Block When It Needs Human Approval
AI agents get stuck waiting for humans. There's a better pattern than blocking - async approval with reminders, escalation, and timeout.
Read post →
Related Cloud features
Related links
Ship your first durable agent — in under 10 minutes.
Free tier. No credit card. Self-host or hosted — your choice.