Aidō Lighthouse
Get a demo
Resources / Muse and Poppy
Standards & Payments · 14 Oct 2026 · 4 min read

Meta Muse, Poppy, and the next personal agent protocol

Meta describes Muse as its personal AI agent. Sierra’s October 9 announcement introduces Poppy, the draft Personal Agent Protocol developed with Meta and other industry partners. The two are connected in a public demonstration: Muse used Poppy to find and work with Rocket’s business agent.

Poppy addresses a practical question: how can a business know when a personal agent is acting for a customer, and what access has that customer approved? Its draft describes discovery through a /.well-known/poppy.json file, agent sessions, customer sign-in and permissioned access. A business can serve the agent through its website, APIs, or its own conversational agent.

Where Poppy fits with commerce protocols

Poppy is broader than shopping. It describes identity, permission, and continuity between a person’s agent and a business. Sierra positions it as complementary to Universal Commerce Protocol (UCP), which focuses on shopping tasks such as product discovery, checkout, and order management. The draft builds on existing technologies including OAuth, JWTs, HTTPS, OpenAPI, and MCP.

The protocol is still a draft being shaped with design partners. Its announcement is a signal of where agent-to-business interactions may be heading; it does not mean every personal agent or merchant currently supports Poppy, or that a published declaration alone proves an interaction will work.

Agent identity does not make a storefront ready

A protocol can help an agent identify itself and request customer-approved access. The merchant still needs usable product information, working site or API paths, and controls that enforce the permissions it grants. A successful real-world journey depends on those pieces working together.

Muse is an example of a personal agent. Poppy is a draft for how personal agents and businesses can work together.

Aidō now includes Meta Muse and Poppy in its scanning and detection coverage. That helps teams see relevant agent signals alongside the storefront conditions that affect access and task completion. A scan reports observed evidence within its scope; it is not proof that every agent can complete every task.

Personal agents are arriving with new ways to identify, ask, and act. Businesses still need to make their systems usable and permission-aware.

Want to know where your store sits?

We scan e-commerce sites and tell you exactly what an AI agent can and cannot do on them. It is genuinely quite nerdy and we love it. Drop us a line if you want us to run the numbers on yours.

Get in touch