Execute Agent-Powered Purchase with Human Approval
The agent securely completes a purchase using a scoped virtual card after obtaining explicit human approval, ensuring no real card details are exposed.
How it works
The agent identifies a purchase need and calls request_purchase_approval with the expected cost, merchant, and reason. It then polls wait_for_purchase_approval until the human approves or denies the request. Upon approval, the agent retrieves the checkout context via get_checkout_context to obtain shipping and billing details. Finally, it completes the transaction using its own checkout capabilities, reporting the result back to the user.
Prompt
Paste this into an agent that has the servers below connected. Adjust the names and dates to your own.
Buy a copy of 'The Pragmatic Programmer' from Amazon for $45. Please approve the purchase request so I can complete the checkout.MCP servers it needs
Source
The vendor page this playbook is drawn from, and the passage that supports it.
Allowance lets agents request scoped human approval, receive limited virtual cards, complete checkout, and report receipts without exposing the human's real card.Allowance MCP documentation