Meta's Muse has moved phone calling out of early access and into its paid tiers. For subscribers, that means an AI agent can dial a business, work through the menu, and wait on hold so you don't have to. The feature has been one of Muse's more practical tricks, and now more people can actually use it.
Meta's Chief AI Officer, Alexandr Wang, announced the rollout on X. His pitch was simple: "let muse take care of all those calls you don't want to make. insurance? cable? DMV? dinner res? not your problem anymore." The feature is available only to paid subscribers in the US.
What you're paying for
Muse has two paid tiers. The "Power" plan costs $20 a month and includes 500 million Muse tokens. The "Maximum" plan costs $100 a month and includes three billion tokens. Phone calls draw from the same token allowance as everything else Muse does, so heavy callers will want to pay attention to the higher tier.
The use cases Meta is pointing to are mundane on purpose: booking reservations, scheduling appointments, and reaching a human agent at a company you pay bills to. Those are the tasks where a person spends twenty minutes in a queue and gets nothing useful. That's where I think an agent like this earns its keep.
How the call actually runs
You start by asking Muse to contact a business for a specific purpose, either by text or voice. From there, the agent builds its own plan. It identifies which business to call, looks up the right phone number, and decides which details to share with whoever answers.
During the call, Muse listens for automated prompts and human greetings, works through menu options, and handles the hold music. Your job is mostly to let it run. That's the real pitch: the agent absorbs the tedious part of the process.
Early users have shared a few success stories. Matt Schlicht, the Moltbook creator and a Meta Superintelligence employee, posted that Muse called Volvo to find out when the next service appointment was open. He said the agent waited on the line for 15 minutes, then got what he needed, and that he now runs multiple calls like this every day. He also shared a recording of the call. The audio was recreated for the video, but the transcript was real, according to his post.
Schlicht's example works because the request is narrow. A single question, a known business, and a clear answer. Open-ended negotiation is a much harder problem, and Muse isn't being marketed for that.
The pushback is already here
Here's the thing: the agent only works if the person on the other end picks up and engages. Many businesses have started hanging up on Muse, citing spam concerns and general frustration with bots. That's a real constraint. If enough companies decide to screen out AI-placed calls, the tool's usefulness shrinks fast.
Muse also isn't alone. Genspark, Instinct Concierge, and CALL-E all offer AI assistants that place calls on their own, and Google has its own calling feature. The category is getting crowded, and the differentiator will probably be how well each agent handles a human who's annoyed to hear from a robot.
Where this goes next
It's a little striking to see a feature that started as an early-access perk become a paid tier option. For anyone who hates waiting on hold with an insurance company, that's a real improvement. I wouldn't hand an agent a hotel booking and walk away, though. The value I see is in the calls that are mostly waiting and reading back account numbers.
The open question is whether this becomes a normal part of how people handle errands, or whether businesses build walls high enough that the agent gets stuck at the front desk. My guess is the answer depends on whether these agents can prove they're worth a human's time on the line.