Meta’s Muse launched September 8, 2026. In the 12 days that followed, it pulled 1.8 million iOS downloads across the U.S. and Canada, according to TechCrunch. ChatGPT managed 1.3 million at the same point in its launch.
But on September 21, the same day those download numbers made headlines, TechCrunch also reported that Amazon had pulled the plug. Muse users got a blunt error: “Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.”
The download story and the Amazon story look separate. They are the same story.
What Muse Is
Meta describes Muse as “a secure, private personal AI agent that proactively helps with people’s goals.” That framing is intentional. Muse is not positioned as a chatbot that answers questions. It is an agent that goes out and does things.
The app runs on Muse Secure VM, a dedicated virtual machine. It is powered by Muse Spark, which Meta calls its “most capable model to date, built for real-world agentic work.”
In practice, Muse can send emails, book travel, open browsers, fill out forms, and negotiate on a user’s behalf. It continues working after the app is closed. It remembers what matters to a user and makes suggestions unprompted. Meta’s example: turning a recipe reel saved on Instagram into a grocery list.
Meta says Muse requests user approval before sensitive actions, including sending emails or making purchases. Conversations are not shared with Meta’s ad systems. Later this year, Meta says, it will introduce Muse Confidential VM, encrypted with a key only the user holds.
Payment goes through Stripe’s Link checkout. Shop Pay integration is coming. Most of what people need is free. Subscription plans exist for users who want to do more.
Muse is available on iOS, Android, and muse.ai, currently in the U.S.
The Download Picture
1.8 million iOS downloads in 12 days, U.S. and Canada only. TechCrunch reports that figure against ChatGPT’s 1.3 million at the same stage.
The raw installs are one thing. The daily active user numbers are another. TechCrunch reports 642,000 daily active users for Muse in the U.S. at that point, versus ChatGPT’s 231,000 at a comparable stage. In a September 25 follow-up, TechCrunch reports Muse averaged 55% day-over-day download growth from September 8 to 17, against 24% for ChatGPT in its first 10 days. Globally, Muse hit 2.8 million total installs across iOS and Android in those first 12 days. The app climbed from number two to number one on the U.S. App Store, overtaking ChatGPT.
One number deserves scrutiny here. TechCrunch reports that over 95% of Muse users are also Facebook users, and 63% are Instagram users. Meta didn’t build an audience for Muse from scratch. It distributed Muse into an existing base.
That context doesn’t shrink the numbers. It explains them. The real question for Muse is retention, not installation. Day one downloads built on social cross-promotion are easy. Keeping users engaged past week one, in a category where ChatGPT and Google Gemini are already habits, is harder.
The Amazon Block
TechCrunch reported on September 21 that Amazon had blocked Muse from completing purchases on Amazon.com. The error message, first surfaced by GeekWire: “Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.”
Amazon cited terms of service. The competitive logic goes further than that.
When a user delegates shopping decisions to an AI agent, that agent responds to the user’s stated objective rather than a sponsored listing. The ad model breaks.
Amazon also has its own foundation models and one of the most popular inference platforms on the internet, TechCrunch notes. The competitive incentive to host a rival AI agent’s traffic runs in the wrong direction entirely.
Walmart and Shopify have integrated with Muse, according to Motley Fool. Amazon has not. Muse can shop for users on what Motley Fool calls Amazon’s two closest e-commerce rivals. It cannot complete purchases on Amazon itself.
What This Signals
Some of Muse’s growth is Meta’s distribution machine. Some of it reflects real appetite for a personal agent, not just a chat interface. TechCrunch cites strong early reviews that praised its “polished design and highly capable AI model,” and Meta announced upcoming features at Meta Connect including video chat with an avatar, computer use on Mac, a dedicated email address, and plans for smart glasses integrations (Source).
The Amazon conflict is a different kind of signal. An AI agent that takes actions, rather than one that answers questions, puts AI companies in direct conflict with the platforms that monetize attention and purchasing behavior. Amazon blocked Muse in the same week Muse was dominating App Store charts. That timing reflects a structural pressure: as AI agents become genuinely capable of acting across the web, the platforms that depend on controlling how users discover and buy things have to decide whether to accommodate them or block them.
Amazon chose to block. Walmart and Shopify chose to integrate.
That split will define how useful AI agents are in practice over the next few years. Not the model quality. Not the interface. The access deals.
For more on how Muse fits into the broader AI assistant field, see our best AI personal assistants roundup for 2026 and our guide to the best ChatGPT alternatives.