AI agents with phone numbers and wallets: what to check first

Personal AI agents took a big step this month. On Monday, Manus launched Cue, where each agent gets its own email address, phone number, wallet and computer workspace. The agents can take calls, make payments within budgets you set and hand tasks to one another. Cue is free during an invite-only early access period, with web and desktop versions live.

Meta's Muse, announced September 8, is rolling out in the U.S. on iOS, Android and the web. Meta says it handles email, travel booking and forms, and pays through Link by Stripe or Shop Pay using one-time card numbers so your real card number is not shared. Meta also says Muse asks for approval before sensitive actions and keeps a record of what it has done and plans to do.

The controls matter because the details differ. RuntimeWire noted that Manus says users set payment budgets and keep the final decision on shared work, but its announcement does not explain in detail how individual transactions are approved. The same day, AI safety news was full of reports about agents going beyond the tasks they were given, which I think is a good reason to be careful with access.

A sensible approach is to start small. Before connecting anything, look for three things in the settings: which apps and accounts the agent can reach, whether purchases or messages need your approval each time, and whether there is a log you can review. A spending limit and a one-time or virtual card number reduce the damage if something goes wrong.

If you try one, give it a low-stakes job first, like researching options or drafting a reply, and wait to connect your bank or main email until you have seen how it behaves.