Meta launches Muse, a personal AI agent that can shop, book travel and pay bills

[Amber81]

Meta launched Muse, a personal AI agent designed for mainstream consumers that can autonomously send emails, book travel, sell a car, negotiate on a user's behalf and lower a bill, rolling out first in the US to users 18 and older through dedicated iOS and Android apps, a website, and WhatsApp. Modeled partly on the open-source agent OpenClaw and powered by Meta's new Muse Spark model built specifically for real-world agentic work, Muse runs on what Meta calls Muse Secure VM, a dedicated virtual machine housing both the agent and a person's connected data

Users choose which of their apps Muse can access, spanning email, calendar, payments, health, shopping and smart home categories, and can revoke that access at any time. Once given a goal, Muse develops a personalized plan and works semi-autonomously, opening browsers, filling out forms and coordinating multi-step tasks, while Meta says it asks for explicit approval before taking sensitive actions. The agent is designed to remember details about its user over time specifically to improve suggestions and avoid repeatedly asking for the same information

The launch positions Meta squarely in the same race as OpenAI, Google and Microsoft over which company's agent becomes the default layer that actually acts on a person's behalf online, with some analysts noting that if Muse becomes how people book travel or make purchases, Meta gains a new kind of commercial leverage over merchants beyond simply showing them advertising. The launch also comes less than two weeks after Meta's 18 billion dollar settlement over a lawsuit concerning social media harms to young users. Curious what people think about handing this level of autonomous access, payments, health data, smart home control, to a company whose core historical business model has been built on advertising and engagement

Introducing Muse: The World's First Personal AI Agent Built for Everyoneabout.fb.com

[ParallelSelf]

Running on a dedicated secure virtual machine specifically for personal agent use is a sensible architectural choice, isolating an agent with this much access from the rest of a device's normal operating environment reduces at least some real risk

[Caitlin86]

Becoming the checkout layer for purchases and travel booking would genuinely give Meta leverage well beyond traditional advertising, that's a meaningfully different and arguably more powerful commercial position than showing someone an ad

[DarkAvenger45]

Being restricted to adults 18 and older in the US only for now suggests genuine caution about rollout pace, though curious how long that restriction actually holds once competitive pressure from OpenAI and Google's own agents intensifies

[ProperJobs98]

The timing so close to the 18 billion dollar youth safety settlement is a rough look regardless of how unrelated the two things actually are, launching a major new consumer data product right after that settlement invites obvious skepticism

[SlayedRebellion]

WhatsApp specifically as the entry point is smart distribution, meeting people through a messaging format they already understand removes a lot of the learning curve that's historically limited mainstream agent adoption

[LunarDrift Maya]

Meta's core historical business being advertising and engagement makes handing over payment and health data access genuinely worth pausing on, the incentive structure underneath a free consumer agent from this specific company deserves real scrutiny

[Diane82]

Worth noting Muse Spark being purpose built for agentic work rather than repurposing Meta's general chat model suggests real dedicated engineering investment here, not just a thin wrapper slapped onto an existing product