What Is Muse?

Muse is Meta's personal AI agent, launched on September 8, 2026. The shortest accurate description: you give it an outcome, and it does the work — planning the steps, using the apps and services you've connected, and continuing in the background after you put your phone down.

That sounds like marketing until you see the difference in practice. Ask a chatbot to "deal with my inbox" and you'll get advice about dealing with your inbox. Ask Muse and it will open your email, triage it, draft the replies, and ask you to approve them before sending. One answers; the other acts.

The one idea: delegation

Everything about Muse follows from a single shift — from conversation to delegation. A chatbot session is a back-and-forth that ends when you stop typing. A Muse task has a goal, a plan, and a life of its own: it can run for minutes or days, check back when something changes, and ask for approval at the moments that matter.

This is why the product feels different from Meta's earlier AI efforts and from the chatbots it competes with. The unit of work isn't the message — it's the job.

What it can actually do

Through connected services (Meta calls them Connectors) and its own browser, Muse can:

• Communicate: read, draft, and — with your approval — send emails and messages
• Book and buy: travel, reservations, and purchases, checking out through Stripe's Link wallet with one-time-use card numbers
• Organize: calendar management, reminders, and scheduled tasks it runs on its own
• Monitor: watch prices, inboxes, or schedules over time and report back when something changes
• Create: documents, PDFs, web pages, trackers, and other artifacts from a brief
• Fill in forms and operate websites directly when there's no API to use

Under the hood it runs on Muse Spark, the model family from Meta's Superintelligence Labs built specifically for agentic work.

Where it runs

As of this writing, Muse is US-only and available in four places: the iOS app, the Android app, the web app at muse.ai, and inside WhatsApp — where you simply message it like a contact. Meta says its AI glasses are next. There is no separate desktop app; the web covers that.

What it is not

Three misconceptions worth clearing up early:

• It's not autonomous in the scary sense. Sensitive actions require your approval, and you set how much it asks about. The safety architecture is a real part of the product, not a footnote.
• It's not magic. It misreads things, especially on unfamiliar websites. The skill is in the brief you write and the review you do — the complete guide is built around exactly that.
• It's not free of cost. There's a free tier, but setup requires a payment card and heavier use moves to paid plans. Pricing explained →

Why it matters now

Muse arrived at an unusual moment: the personal-agent category is brand new, and there's almost no established wisdom about how to use these things well. The people who learn the delegation pattern early — clear briefs, tight approvals, one job at a time — will get dramatically more out of it than people who treat it like a chatbot with extra steps.

That's what this publication is for. Start with getting set up, then give it one real job.

Quick answers