Meta Muse AI Agent Launches in the US: Apps, WhatsApp, and Paid Tiers

Meta just shipped Muse, a personal AI agent that is supposed to do more than chat. On September 8, 2026, the company opened a US rollout that connects email, calendars, payments, WhatsApp, and more—while asking users to trust Meta with a deeper slice of everyday life than a social feed ever required.

What Muse actually does

According to TechCrunch’s launch coverage, Muse is built to send emails, book travel, lower bills, fill forms, turn recipe Reels into grocery lists, send party invites, and make purchases. Checkout leans on Link by Stripe, with Shop Pay and 1Password integrations described as coming soon. That is the agentic pitch in plain language: not another answer box, but a helper that can finish multi-step chores after you leave the app.

Users opt in service by service. Meta ships built-in connectors and, when a public API exists, can wire credentials you provide. If there is no API, Muse can fall back to a browser session inside its cloud environment. The agent is powered by Meta’s Muse Spark model and keeps working in the background once you hand off a task.

Where it lives—and what it costs

Muse starts on the web at muse.ai, in iOS and Android apps, and inside WhatsApp chats. Meta says AI glasses support is coming soon. The product is free for light use, but Meta requires a payment card up front because paid plans kick in when you burn through free capacity. TechCrunch cites two launch subscriptions: Power at $20/month and Maximum at $100/month, with an in-app usage meter and warnings before free quota runs out. Meta expects most people to stay on the free tier.

Security claims vs. trust baggage

Meta says Muse runs in a dedicated Muse Secure VM with its own browser, supervised by a separate Sentinel agent that stays isolated at the system level. The company claims Muse cannot see passwords or payment methods and does not share conversations with Meta’s ads systems. Those are Meta’s statements—security researchers will need to pressure-test the technical post Meta published alongside the launch.

Reuters’ same-day report also noted that internal tests had shown the product stalling and raising concerns about unauthorized exposure of sensitive data—context that matters if you are about to connect a bank or health app. TechCrunch separately frames the launch against Meta’s long privacy settlement history, including an $18 billion multistate social-media settlement less than two weeks earlier. Rocks’ takeaway is simple: utility is real on paper; trust is the product risk.

Why Rocks cares

If agentic AI is the next interface after chatbots, Muse is Meta’s mass-market swing—and it lands on Rocks’ tech planet because it touches phones, WhatsApp, payments, and smart-home connectors in one release. Try the free tier only with the least-sensitive connectors first, read the Secure VM claims critically, and treat the $20/$100 tiers as a usage tax you might hit faster than Meta’s marketing implies.

Sources: TechCrunch, Reuters, The Verge.