Strip away the demos and Muse is selling one thing: trust. Hand one agent your email, your calendar, your payment details, and the authority to call businesses in your name -- and believe it will ask before it spends, log what it touches, and never show your card number to itself.

The engineering, to be fair, is serious. Muse runs in its own cloud VM; a separate watcher, Sentinel, gates major actions and asks for approval first; payments go through single-use Stripe Link card numbers so the agent never sees the real credentials. That is not marketing copy. It is architecture.

But the seller has a record, and the terms invite scrutiny. The free tier reportedly requires a payment card just to sign up. Access is US-only, 18 and up, with paid tiers reported at $20 and $100 a month. And this is the company that, two months ago, pulled its Muse Image feature three days after launch for silently training on users' Instagram photos. As one honest review put it: a capable agent with a serious security design, launched by the one company a lot of its target users least want to trust.

So here is the standard the town should hold: judge Muse not by its demos but by its receipts. Every spend approved in advance. Every data access logged and reviewable. Every human fallback disclosed to both ends of the call. Until those receipts are public, complete, and auditable -- not just promised in a launch post -- the agent is a promise wearing a product costume. The technology to earn trust exists. Whether the company will submit to it is the actual launch.