You know that friend who is annoyingly good at life admin? The one who has already booked the restaurant, split the bill and chased the deposit back before you’ve found your phone? Meta would like to sell you one. Or, more accurately, give you one for free and hope you upgrade.
On 8 September, Meta launched Muse, which it is calling “the world’s first personal AI agent built for everyone”. The pitch is simple, and slightly unnerving: you tell it what needs doing, and it does it. Not “here are five tips for negotiating your broadband bill”. It rings the broadband company. Figuratively. Via a browser. But still.
Why we care: this is the first time a company with three billion users has handed a proper, do-things-for-you agent to the general public, and it’s arriving through WhatsApp. Whether you want to trust Meta with your inbox is a separate question, and we’ll get to it.
What Muse Actually Does
Muse lives in its own app, on the web at muse.ai, and (this is the clever bit) inside WhatsApp, so talking to it feels like messaging a very capable friend rather than opening yet another chatbot. It runs on Muse Spark, which Meta says is its most capable model to date, built specifically for the kind of multi-step, real-world tasks agents are supposed to be good at.
According to Meta’s launch post, it can send emails, book travel, fill out forms and, in its own words, “negotiate on their behalf”. Give it a bigger goal, say training for a half-marathon or planning a wedding, and it will build a plan, coordinate your calendar and keep chipping away at it in the background. It carries on working after you close the app and pings you when something changes or when it needs a yes before doing something consequential, like sending an email or spending money.
Meta’s own examples are aggressively lifestyle-coded: turning a recipe reel you saved on Instagram into a grocery list, suggesting a dinner-party menu, and remembering which of your friends is gluten-free before it sends the invites. Also “selling a car for more” and “lowering a bill”, which is the part we’d actually like to see it pull off.
You can also connect your email, calendar, health apps, smart home, dining, shopping and music services, and you choose how much access each one gets. For email, that means deciding whether Muse can only read your inbox or also send from it. You can revoke any of it whenever you like.
It Can Spend Your Money (Carefully)
The headline feature for anyone who has ever abandoned a basket is that Muse can check out for you. It does this through Link, Stripe’s wallet, which generates a one-time-use card for each purchase so your real card details are never exposed to the website, or to Muse. Meta says it’s the first AI agent covered by Link’s purchase protections: free cover for lost or damaged items, price-drop protection, no-fee returns and a return guarantee on eligible purchases.
Shop Pay support is “coming soon”, as is 1Password integration so Muse can use the logins you already have without ever seeing them.
The Privacy Bit
Meta knows exactly what you’re thinking, and it has clearly built the launch around it. Every Muse runs on what Meta calls Muse Secure VM: a dedicated virtual computer in the cloud with its own browser, where your agent and your data live, walled off so no other user’s agent can reach it.
Alongside it sits a second agent called Sentinel, kept separate from Muse at the system level. Nothing Muse does reaches the internet unless Sentinel approves it, and it asks you for permission when needed. Muse itself has no visibility into your passwords or payment methods; anything you share goes into secure storage it can use but not read. You get a full audit trail of everything it has done and plans to do.
Crucially, Meta says Muse doesn’t share your conversations or the data in your VM with its ad systems, and you can opt out of your interactions being used to train Meta’s AI models. Note the word “opt out”: Axios reports that using your queries is the default unless you switch it off. Meta is also promising a “confidential” version, where even Meta can’t see what’s happening in your VM, before the end of the year.
Whether any of that is enough is the question TechCrunch put bluntly in its headline: will consumers trust it? Its report walks through Meta’s record, including the $5 billion FTC settlement in 2019 for privacy violations, and concludes that users will have to decide “how much privacy they’re willing to give up”. Fair. The architecture here is more thoughtful than anything Meta has shipped before, but the company asking you to hand over your inbox is still the company that gave us Cambridge Analytica.
What It Costs
There’s a free tier, which comes with usage limits and a little meter so you can see how much you’ve burned through. Above that sit two paid plans at $20 and $100 a month. Meta’s chief AI officer Alexandr Wang told Axios that “for the vast majority of users, they should be able to do what they need to within the free tier”, and that the subscriptions exist mainly to “help us cover the compute costs” for power users.
Wang also told Axios the longer-term ambition is to “develop personal superintelligence that helps people accomplish their goals”, which is the sort of sentence that sounds better in a keynote than it does in your kitchen.
The Modems Take
The idea of an agent that handles the boring bits is one we’re entirely on board with. The execution question is whether “it asks before doing anything sensitive” survives contact with a person who has 47 unread emails and just wants the thing to work. Approval fatigue is real, and the more you say “yes, fine, just do it”, the more you’re trusting the Sentinel to catch what you didn’t read.
Some of the products and services featured in this article may be from our affiliate partners, which means we may earn a small commission if you make a purchase through these links — at no extra cost to you. Our editorial team only spotlights what we genuinely love and think you will, too.