Draft poster · the finished cut will appear here
What happened
Meta introduced Muse as a personal AI agent that can take action on goals. Meta says Muse runs in a dedicated virtual machine or cloud computer with its own browser, and can keep working after the app closes. The company describes it opening browsers, filling forms, negotiating, sending email with approval, and making purchases.
How it works
The payment boundary is the notable part. Meta says Muse checks out with Link by Stripe, whose wallet for agents generates a one-time-use card so the user’s real card details stay hidden. Meta says eligible purchases also receive Link’s purchase protections. A separate Sentinel agent controls internet access, while Muse requests permission for sensitive actions and records an audit trail.
That architecture puts execution, credentials, policy, and recovery around the language model. The model can operate a browser, but it does not need to receive the user’s ordinary card number. Meta says Muse is rolling out in the United States; Shop Pay and 1Password support are planned.
Why it matters
Agentic commerce becomes credible only when the payment token, approval rule, execution environment, and record of what happened are designed together. A chatbot with a card is a very different system from an agent with bounded credentials and an explicit checkpoint.
What we don’t know yet
Security, privacy, purchase protection, and first-to-market positioning are Meta or Stripe product claims, not an independent audit. Rollout limits and approvals still matter, and planned integrations are not available features.
Source: Meta Newsroom, Meta, September 8, 2026 - original announcement