How to use OpenClaw for dinner and multileg flights
OpenClaw’s Alex proposed dinner options and booked after approval, then handled reported multileg flights; the source leaves flight payment checkpoints unclear.
linkedin.com
Original prompt
Shared by the source author. See source notes for context.
Dinner for 2 in West Hollywood tonight.
Reach people exploring AI agents
See placements →What happened
The author reports a confirmed dinner reservation and booked multileg flights; exact flight prices and authorization details are not disclosed.
Reported by Amanda Merrifield, MBA — LinkedInHow the agent did it
- Merrifield configured a dedicated Mac mini assistant and supplied personal preferences and connected context.
- Alex proposed three dinner options, requested approval and produced a reported reservation confirmation.
- The author watched a multileg flight workflow and reports completed bookings using her travel preferences.
Where a person stepped in
Dinner and shopping approvals are explicit; the author watched the flight browser run. The post does not publish all guardrail settings or flight payment checkpoints.
Amanda Merrifield built an OpenClaw assistant called Alex on a dedicated Mac mini and used iMessage to delegate personal arrangements. Her account describes completed reservation and flight workflows, rather than only a proposed assistant design.
For dinner in West Hollywood, she supplied the party size and location. Alex returned three recommendations with ratings, cuisines, times and links, then asked whether to book. Merrifield approved, and reports receiving the confirmation in her inbox shortly afterward.
A more complicated task involved multileg flights connected to Barbados travel. The assistant worked from her preferences, calendar and rewards memberships while she watched it operate the browser. She reports that the flights were booked after about fifteen minutes. The post does not expose the full route, price or individual payment handoffs, so those details remain outside the evidence.
The same account describes an Amazon reorder that used an earlier purchase and an approval before proceeding. That supports the documented approval pattern, but does not prove every flight action used an identical checkpoint. Merrifield mentions guardrails without publishing their configuration. The outcome is her reported use of an operating personal assistant, with specific dinner approval evidence and less detailed flight authorization evidence.
Prompt
Original prompt
Source: LinkedIn · Amanda Merrifield, MBA
Frequently asked questions
How did the owner communicate with Alex?
She used iMessage with an OpenClaw assistant running on a dedicated Mac mini.
Was dinner booked without a choice from the human?
No. Alex supplied options and asked whether to book before Merrifield approved.
Are all flight approval steps documented?
No. The post reports watching the run and successful booking but does not expose each payment or authorization step.
More details
About this account
Merrifield configured a dedicated Mac mini assistant and supplied personal preferences and connected context. Alex proposed three dinner options, requested approval and produced a reported reservation confirmation.
Primary post and main JSON-LD date opened. Original dinner prompt excerpt. Flight completion is author-reported; undisclosed guardrails and payment handoffs are not presented as verified controls. Original prompt excerpt.
Credit: Amanda Merrifield, MBA — LinkedIn
Source published 24 Feb 2026
Source checked 14 Oct 2026
Tools and topics
OpenClaw · Mac mini · iMessage · Web browser
dinner reservation · flight preferences · approval messages
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.