## TL;DR
An agent waiting indefinitely on a user holds state and blocks follow-ups. Set a response timeout with a nudge, then auto-park the conversation with a summary so it can resume cleanly later.

## The query
```text
agent timed out waiting for the user to respond: helpdesk conversation timeout design
```

## Use this when
- agent sessions hang waiting for users
- stale sessions pile up
- users return to dead conversations

## Not for
- API or tool call timeouts
- agent-to-agent handoff delays
- scheduled job timeouts

## Steps
1. Define a response timeout per conversation type, such as 15 minutes for chats and 24 hours for email. Expected output: timeouts are explicit policy.
2. Send one nudge before the timeout, restating what is needed. Expected output: the user gets a clear reminder.
3. On timeout, park the conversation with a summary of state and next steps. Expected output: the session closes cleanly with context saved.
4. Allow resume from the parked summary when the user returns. Expected output: the conversation continues without repeating questions.
5. Monitor parked-conversation rates to tune the timeouts. Expected output: timeouts match real user behavior.

## Applies to
Conversational helpdesk agents, chat and email channels, any agent framework with session state.

## Variant phrasings
### User replies after parking
Resume from the summary; confirm the facts are still current first.

### Timeouts too aggressive for executives
Per-segment timeouts beat one global value.

## Why it happens
Users get distracted; agents do not. Without a timeout the session leaks resources and the eventual reply lands in a confused context.

## Edge cases
- P1 incidents need shorter timeouts and escalation, not parking.
- The parked summary must include what the agent already tried.
- Nudge wording matters; blame the process, not the user.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_i2aKNt7QzBowpuy7Z8EfGQ