Meta launches Muse, a personal AI agent that books, buys and negotiates for you

Meta has launched Muse, a personal AI agent that doesn’t just answer questions but does the work. Released in the US on September 8, 2026, Muse can send emails, book travel, fill in forms, negotiate and buy things on your behalf, and it keeps going after you close the app. It runs on its own secure computer in the cloud, and you talk to it the way you’d message a person — in the Muse app or directly in WhatsApp.

What Muse does

You tell Muse what needs doing, from a single errand to a long-term goal. For bigger goals, it helps you build a plan, organises your time and resources and then moves the work forward on its own. It can open a browser, fill in forms and negotiate — Meta’s examples include selling a car for more, lowering a bill and adjusting a training plan as the rest of your life changes.

Muse also remembers what matters to you. It can make suggestions unprompted, act on something you mentioned once, turn a recipe reel you saved on Instagram into a grocery list, suggest a dinner party menu and remember your friends’ dietary restrictions before it sends the invitations.

For longer jobs, Muse keeps working in the background and comes back when something changes or when it needs your approval — for example, before it sends an email or makes a purchase.

How Muse keeps your data separate

Meta built a dedicated computer for every user, called Muse Secure VM:

• Your own cloud computer. Muse runs on a contained virtual machine with its own browser. Your data and the credentials for services you connect are stored there, and no one else’s agent can reach it.
• A second agent as gatekeeper. A separate Sentinel agent runs on the same machine, kept apart from Muse. Nothing Muse does reaches the internet unless the Sentinel approves it, and it asks you when needed.
• Hidden credentials. Muse can use passwords and payment methods you share without seeing them, including passwords you type into its browser yourself.
• Approval and an audit trail. Muse checks with you before sensitive actions and shows a full record of everything it has done and plans to do.
• You choose the access. You pick which apps Muse connects to and what it can do — for email, whether it can only read or also send. You can disconnect services at any time.

Meta says Muse doesn’t share your conversations or VM data with its ad systems, that you can opt out of your interactions being used to train its models, and that you can tell Muse to forget things. Later this year, Muse Confidential VM will encrypt the whole machine, including your conversations, with a key only you hold, so Meta can’t access it.

Paying with Muse

Muse checks out with Link by Stripe, and Meta says it’s the first AI agent covered by Link’s purchase protections, including cover for damaged or lost items, price drops and no-fee returns on eligible purchases. Link generates a one-time-use card, so your real card details stay hidden. Shop Pay and 1Password support are coming soon.

Price and availability

• Where: rolling out in the US on iOS, Android and the web at muse.ai, and through WhatsApp. Meta says it’s coming soon to AI glasses; see our guide to the best AI smart glasses.
• Free plan: Meta says Muse is free for most of what people need.
• Paid plans: TechCrunch reported two plans at launch, Power at $20 a month and Maximum at $100 a month, both offering more Muse usage for handing off everyday tasks.
• Model: Muse runs on Muse Spark, the model family from Meta Superintelligence Labs, which Meta calls its most capable model for real-world agentic work.

Update, September 20: Muse for Mac, Canada and new connectors

Meta’s Muse account said on X that the team shipped four things this week: Muse for Mac, an expansion to Canada on iOS and the web with Android to follow, new connectors with Granola (see What is Granola AI?) and Notion, and a Muse developer platform. The post gives no further detail on the developer platform.

An early hit on iPhone, a slower start elsewhere

Two days after launch, TechCrunch reported that Muse had been downloaded more than 83,000 times on iOS in the US, according to Sensor Tower, reaching No. 2 on the App Store’s top charts. The start was slower than some of Meta’s previous launches: Threads was downloaded more than 4.3 million times in the US on its first day, and the Meta AI app 108,000 times at its debut. On Google Play, Muse ranked No. 338 in the Productivity category.

The trust question

An agent that reads your email and spends your money asks for a lot of trust, and TechCrunch pointed to Meta’s record: a 2011 settlement with the US Federal Trade Commission over charges that it deceived users about privacy, a $5 billion FTC settlement in 2019, and the 2019 discovery that it had stored users’ passwords in readable form. Meta’s answer is the architecture above — an isolated machine, a gatekeeper agent, hidden credentials and, later, encryption Meta itself can’t unlock.

If you try Muse, start with low-stakes tasks, connect only the apps it needs and read each approval request carefully before you confirm an email or a purchase.

How Muse compares

Muse joins a fast-growing group of consumer agents that take action on their own: OpenAI’s ChatGPT Work, Perplexity Computer, Claude Cowork and SpaceXAI’s Grok Bot among them; see our guide to AI agents explained for how this category works more broadly. Muse’s pitch is reach — WhatsApp and Meta’s apps — plus the security design around payments and passwords. Muse can also send email once you approve it; see how the alternatives compare in our roundup of the best AI email assistants. Meta also has a coding agent, Muse Code, in beta, covered in our guide to the best AI coding assistants. For help choosing an assistant, see our guide to the best AI personal assistant. At Connect on September 23, Meta said its AI glasses connect to Muse; see Meta Connect 2026: everything announced.