Meta Muse Hits iPad With 6.6M Installs as Agent Identity Standard Takes Shape

Meta’s Muse assistant arrived on iPad on October 7, 2026, roughly a month after its September 8 debut on iOS and Android — a multidplatform rollout pace it took Instagram some 15 years to match. The app has already passed 6.6 million installs, according to estimates from market intelligence firm Sensor Tower, and Meta used the tablet launch to widen its connector catalogue while backing an open standard that would let personal AI agents identify themselves to the businesses they visit.

• Faster than expected: Muse reached iPad in about four weeks, a signal that Meta is prioritising installed base over product polish in the consumer agent race.
• Connectors go enterprise and retail: Asana, Canva, Dropbox, Figma, QuickBooks, GitHub, Klaviyo, Zoom, Notion and Granola join the productivity side, while Best Buy, Gap, Sephora, Walmart and Wayfair bring commerce.
• Identity is the new battleground: The Personal Agent Protocol, unveiled with Sierra on October 6, 2026, tackles how agents authenticate with merchants — but its specification has not been published yet.

A Bigger Canvas and a Longer Connector List

The iPad build leans into the larger display: while the agent works through a task, it can surface its reasoning process and the source documents side by side. For professional use this matters more than it does for casual queries. An analyst asking an agent to review a financial report needs the working and the underlying data on screen at once, not sequentially.

The connector roster reads less like a chatbot feature list and more like an enterprise automation platform. Notion and Granola cover notes and knowledge management, while Klaviyo points to high-volume email marketing, suggesting Meta wants Muse operating on live campaign data. The pairing of Figma and GitHub is the most consequential for product teams, since design files and repositories usually sit in separate silos; whether the agent can reason across both at once will determine how much that integration is actually worth.

On the commerce side, Wayfair and Sephora imply guidance for high-consideration purchases such as furniture and cosmetics, while Best Buy targets electronics, where comparing specifications between products is precisely the kind of task an agent should handle better than a conventional search box. Meta has also added connectors for its own ad accounts, and a Windows version is reported to be in development. Muse is now reachable across phones, Mac, iPad, the web and WhatsApp.

When the Storefront Says No

The optimistic framing runs into a practical obstacle. Amazon began blocking Muse from its retail site on September 20, 2026, with a spokesperson arguing that apps buying on customers’ behalf should operate openly and respect the choices of the merchants they shop. Adweek relayed a further Amazon claim that Muse does not identify itself — an allegation that has not been independently verified. Shopify took the opposite route, partnering with Meta so that Muse purchases can run through Shop Pay.

Meta’s response to the friction was blunt: “Turning away a personal agent means turning away the customer behind it.” That single line frames the company’s entire argument — blocking an agent now means turning away a paying human.

Inside the Personal Agent Protocol

Announced on October 6, 2026 at the Sierra Summit in San Francisco by Sierra co-founders Bret Taylor and Clay Bavor, the Personal Agent Protocol (PAP) arrived with Walmart, Shopify, Stripe, Genesys, Instinct and Rocket on the partner list — and no specification attached. Sierra says the v0.1 text will follow later in the month.

The design starts at a company’s own website. An agent discovers what a business offers and can begin as a guest to check stock or a returns policy. For account tasks the customer signs in on the merchant’s page or with credentials already stored by the agent, then chooses read-only or write access. Sessions run on OAuth and persist across channels, so a pre-login question and a post-login order change belong to the same visit. Businesses choose between ordinary web pages, APIs built on MCP and OpenAPI, or their own customer-facing agent. Payments are listed only as a possible future extension, with no dates.

Tony Bates, CEO of Genesys, framed the gap the protocol addresses: “Personal AI is creating a new front door to the enterprise. Brands need a trusted way to know who an AI agent represents, what it’s authorized to do, its intent, and how to work with it securely.”

David Singleton, VP of Engineering at Meta Superintelligence Labs, described the ambition as infrastructure: “We’re defining rails that we hope personal agents and business agents can run over for the future,” comparing the protocol’s function to email — a shared transport layer rather than a product. Taylor has said he expects OpenAI and Anthropic to sign on, according to reports, though neither has committed publicly.

The overlap with rival efforts is striking. Decagon, itself listed as a Meta partner, published its own Personal Agent Consent & Trust Protocol on the same day, built on the Agent2Agent protocol and OAuth 2.0 and co-developed with Instinct. Meanwhile Stripe and Shopify appear in PAP, Visa’s Trusted Agent Protocol, OpenAI’s Agentic Commerce Protocol and Google’s Universal Commerce Protocol, and Walmart in both PAP and Google’s effort. That multi-homing looks less like confusion than deliberate hedging.

The structural bet is that identity and permissions should come before payments: a checkout standard is only as useful as a merchant’s confidence about who is asking and what they may do. The catch is that none of this can be tested until the specification exists. Muse’s six million-plus installs give Meta the distribution to push that standard, but whether enterprises adopt the assistant as corporate infrastructure rather than a personal app depends on how thoroughly those connectors mature — and on whether the merchants being asked to trust agents decide that the trade is worth making.