Muse AI by Meta: Personal AI Agent Explained (2026)

Meta’s Muse isn’t another chatbot — it’s a personal AI agent that books appointments, fills out forms, shops, and even calls businesses on your behalf. Here’s what it actually does, what it costs, where it’s available, and the trust questions worth asking before you connect your accounts.

September 29, 2026  ·  12 min read

Table of Contents

• What Muse actually is (and isn’t)
• What Muse can do for you
• Muse pricing: free, $20 and $100 tiers
• Where you can use it (and who can’t yet)
• Muse vs the other AI agents
• The trust question: should an agent spend your money?
• Who Muse is for — and who should wait
• FAQ

Most AI assistants in 2026 still work the same way: you ask, they answer, and you do the actual work. Draft the email, book the flight, call the plumber — the last mile is always yours. Meta’s Muse, launched on September 8, 2026, is built on the opposite bet: that the assistant should do the doing.

Meta calls it “the world’s first personal AI agent built for everyone.” That’s marketing, but the product underneath is genuinely different from a chatbot. This guide explains what Muse is, what it can verifiably do today, what it costs, and where the real caveats are.

Research note: Muse launched US-only and we haven’t tested it hands-on. This is a research-based overview built from Meta’s official announcements and independent reporting (Reuters, TechCrunch, Axios, CNBC, CNET), linked in Sources below — consistent with how we review AI tools.

What Muse actually is (and isn’t)

Muse is a personal AI agent: software that takes a goal you state in plain language and works through the multi-step task itself, across your connected apps and the web, with minimal supervision. Tell it to plan a weekend trip, and it doesn’t just suggest an itinerary — it checks your calendar, compares flights, and books.

Meta’s own description, cited across launch coverage: “Once a person shares a goal with Muse, it helps them develop a personalized plan and coordinate their time and resources, then advances the work on its own.” The company adds that Muse “can open a browser, fill out forms, and negotiate on their behalf.”

Under the hood, two things matter:

• Muse Secure VM. Each user gets an isolated virtual machine in Meta’s cloud with its own browser. You can watch the agent work in a visible browser window rather than trusting it blindly — booking, form-filling and navigating sites as you would.
• Muse Spark lineage. Muse runs on Meta’s Muse Spark model family, the same agentic model line behind Meta’s developer tools, overseen by chief AI officer Alexandr Wang.

Mark Zuckerberg framed the ambition bluntly: “Everyone will have an exceptionally capable personal agent that understands you, your goals, and everything you care about. Your agent will work 24/7 on your behalf to improve your relationships, health, career, finances, home management, hobbies, and more.” Big claim — but it tells you exactly what Meta is aiming at.

What Muse can do for you

According to Meta and launch reporting, Muse connects to at least six areas of your digital life — email, calendar, payments, health, shopping, and smart home — plus messaging surfaces like WhatsApp and Instagram. Practical things it can do:

• Run errands end-to-end. Send emails, schedule meetings, shop for products, and book travel without you touching each step.
• Browse and fill forms. Inside its secure VM browser it can navigate websites, complete forms, and interact with customer-service systems.
• Call businesses. A phone-calling feature lets Muse dial US businesses and talk to whoever answers — booking a haircut, checking whether a store has an item in stock, or collecting quotes from contractors. (More on the controversy around this below.)
• Work in the background. Give it a goal, and it keeps working while you do other things, checking back when it’s done. You can also summon it by name.
• Build its own tools. If a task needs a capability that doesn’t exist yet, Meta says Muse can create the tool it needs — one of the more genuinely “agentic” claims in the current market.
• Remember context. It retains information from earlier conversations and applies it to longer-term goals, reminders, and recurring tasks.

Muse pricing: free, $20 and $100 tiers

Muse launched with three tiers — free, Power at $20/month, and Maximum at $100/month — reported by TechCrunch and Axios and confirmed directly to CNBC by Meta. Unusually, the tiers differ in usage allowance, not features: paying more buys headroom, not new capabilities.

The free tier is genuinely generous — 100 million tokens a week is far beyond casual use — which signals Meta is prioritizing adoption over near-term revenue. The jump to $100/month puts the top tier closer to business productivity software than to a typical $20 AI subscription, and it’s aimed at people running the agent continuously across shopping, scheduling, and household admin.

Pricing changes frequently. Check the official pricing page for the latest price.

Where you can use it (and who can’t yet)

• Platforms: dedicated iOS and Android apps, the web at muse.ai, and WhatsApp. A Mac app followed on September 18.
• Region: US at launch (September 8); Canada added September 18.
• Age: 18 and older.
• The catch: Meta requires a valid payment card on file before you can use Muse at all — even the free tier. That functions as age/identity verification and anti-abuse protection, but it’s a higher bar than most free AI products, which ask only for an email.
• Coming soon: at Meta Connect 2026 (September 23), Meta said it’s bringing Muse to its AI glasses — hands-free workouts guidance, meal logging, nutrition insights, buying products you see, booking appointments, or tracking down a missing part while you’re driving or cooking, all without pulling out your phone.

Muse reportedly topped US app download charts in its first two weeks — so despite the card requirement, launch demand was strong.

Muse vs the other AI agents

Muse isn’t launching into an empty field. OpenAI, Google, and Perplexity all shipped agent products in 2026. The honest difference is distribution and interface:

Meta’s bet is that you shouldn’t have to change how you browse the web to use an agent — you message it the way you’d message a person, inside WhatsApp, where billions of people already are. Whether that beats a browser-native agent depends on the tasks: browser agents see the page you’re on; Muse sees your messages, calendar, and connected apps. For a deeper head-to-head on two leading assistants, see our ChatGPT vs Claude comparison.

The trust question: should an agent spend your money?

This is the central question around every 2026 agent, and TechCrunch framed its Muse coverage around it directly: will consumers trust an AI enough to let it act autonomously with their accounts and money?

Meta’s answer is a permission model: Muse asks for explicit approval before sensitive actions — specifically before sending a message on your behalf or spending money. You also control which data and accounts it can access in the first place. That’s the same safety pattern other agent products converged on this year: autonomy for low-stakes work, a hard human stop before anything irreversible.

But two caveats deserve honest weight:

• The “human concierge” report. Shortly after launch, Reuters reported — from internal Meta posts — that the company had been testing human contractors quietly handling some of Muse’s phone calls, and that some employees raised privacy concerns, warning that sensitive information could be shared unintentionally with call-center contractors. One employee wrote: “It’s baffling to me why we think this feature is worth the risk… We are one bug away from unnecessary information being leaked to human callers.” Meta hasn’t disputed the test; treat the phone feature as partly human-assisted for now.
• Approval gates only work if you read them. An agent that asks permission 40 times a day trains you to tap “approve” without reading. If you connect payments, keep notifications on and review what Muse did — the visible browser window exists precisely so you can watch.

On the positive side, Reuters also reports Meta plans an encrypted Muse mode to further wall off sensitive data. Directionally good; verify it exists before relying on it.

Who Muse is for — and who should wait

Muse is worth trying if you: live in the US, are 18+, juggle lots of small errands (appointments, reservations, comparison shopping), and already live in WhatsApp. The free tier’s allowance is large enough that most people will never hit a paywall.

Wait or skip if you: are outside the US and Canada (the only two countries so far), are uncomfortable putting a payment card on file for a free product, or want an agent primarily for coding and development — that’s what Muse Spark and Muse Code are for, not the consumer agent. And if the idea of an AI calling businesses in your name makes you uneasy, the Reuters reporting gives you legitimate reason to hold off until the human-concierge question is answered publicly.

FAQ

What is Muse AI by Meta?

Muse is Meta’s personal AI agent, launched September 8, 2026. Unlike a chatbot that only answers questions, Muse executes multi-step tasks — booking, shopping, scheduling, form-filling, and phone calls — across your connected apps and the web, inside an isolated secure virtual machine.

How much does Muse cost?

There are three tiers: Free (reported up to 100M tokens/week), Power at $20/month (500M tokens/week), and Maximum at $100/month (3B tokens/week). The paid tiers buy more usage headroom, not different features. Pricing changes frequently — check the official pricing page for the latest price.

Is Muse available outside the US?

It launched US-only on September 8, with Canada added on September 18 – those are the only countries so far. It’s 18+, and Meta requires a Meta account — check Meta's official pages for current signup requirements.

How is Muse different from ChatGPT or Meta AI?

Chatbots answer; Muse acts. A chatbot drafts the email for you to send, while Muse can send it, check your calendar first, and follow up — asking your approval before anything sensitive. Meta AI (the assistant inside Facebook/Instagram/WhatsApp) answers questions; Muse is the separate agent product that completes tasks.

Can Muse really make phone calls?

Yes — it can call US businesses to book appointments, check stock, or collect quotes. However, Reuters reported in September 2026 that Meta quietly used human contractors for some of these calls during testing, which raised employee privacy concerns. Treat the feature as partly human-assisted until Meta clarifies.

Is it safe to let Muse access my email and payments?

Meta’s permission model requires your explicit approval before Muse sends messages or spends money, and you control which accounts it can see. That’s the industry-standard pattern, but approval fatigue is real — review what it does, especially early on. An encrypted mode is reportedly planned.

Conclusion

Muse is the clearest consumer test yet of the agent thesis: not a smarter chatbot, but software that does the errands. The free tier is generous, the WhatsApp distribution is clever, and the Secure VM + approval-gate design is the right safety shape. The open questions — the human-concierge phone calls, limited availability (US and Canada so far), and whether approval gates survive contact with 40 notifications a day — are exactly what to watch over the next few months.

If you try it, start small: connect the calendar, watch the approvals, and let it earn the payments connection.

Sources

• Meta Newsroom — “Introducing Muse for Small Business,” official announcement (September 2026). about.fb.com
• Reuters — “Exclusive: Meta testing a ‘human concierge’ for its new personal AI agent, Muse” (September 22, 2026). Channel NewsAsia (Reuters)
• Launch pricing roundup citing TechCrunch, Axios, CNBC, Reuters and CNET reporting — free tier, Power at $20/month, Maximum at $100/month. tech-insider.org

Pricing and availability checked October 3, 2026; details may change — check Meta’s official pages for the latest.