Meta's New AI Doesn't Just Chat. It Books Your Flights and Sells Your Car

Keypoints:

• Meta launched Muse, a personal AI agent, on September 8, available to US users 18 and older
• It runs inside a dedicated secure virtual machine with its own visible browser window
• Muse can send emails, book travel, negotiate bills, fill out forms, and complete purchases
• It works standalone or directly inside WhatsApp, with pricing tiers at $20 and $100
• The launch came less than two weeks after Meta's $18 billion multistate consumer-harm settlement

Meta introduced Muse, a personal AI agent built to complete real-world tasks rather than answer questions, marking the company's biggest consumer AI bet to date. Instead of chatting back and forth, Muse takes a goal, a bill to negotiate, a trip to book, a form to fill out, and works through it inside its own secure virtual machine, with a visible browser window so users can watch it operate rather than trust it blindly.

The task list runs wide. Meta says the agent can send emails, manage a calendar, buy movie tickets, schedule appointments like tennis lessons, fill out a permission slip for a class field trip, and even sell a car, connecting across a user's email, payments, health services, shopping, and smart home accounts to get there. Meta's own framing described it as "the world's first personal AI agent built for everyone," language that reads more like a category claim than a routine feature update.

The product is powered by Muse Spark 1.3, the same model family behind an agentic coding tool Meta shipped earlier in September, developed under the company's chief AI officer, Alexandr Wang. 

It rolls out first in the US on iOS, Android, and the web, with WhatsApp integration built in and AI glasses support coming soon, priced at $20 and $100 monthly tiers, with a more locked-down "Confidential" mode planned for later in the year. It's expected to compete directly with rival agent products from OpenAI and other labs.

The timing raises its own question. Muse launched less than two weeks after Meta agreed to an $18 billion multistate settlement over social media's harms to users, and handing an AI agent access to email, payments, and health information asks for a different level of trust than a news feed ever did. 

Tech analyst Jack Gold framed the urgency behind the launch directly: Meta needs to stake out a position in agentic AI before Microsoft and Google absorb most of the users already inside Meta's own apps. Meta is betting the visible-browser design and granular permission controls are enough to clear that trust bar, in a race against OpenAI, Google, and Microsoft over who builds the AI agent people actually let run their errands.