Poppy 1.0: Power-User Guide to Sierra’s Personal Agent Protocol

Last updated: November 15, 2026

Personal Agent Protocol 1.0, or Poppy, is an open standard Sierra released on November 13, 2026. It lets a personal AI agent find a company, sign the user in through OAuth and work through the company’s APIs, website or agent. Meta’s Muse now checks for Poppy on every site, and every Shopify store publishes a Poppy file.

This guide explains what changed in version 1.0, how the protocol works, how a company can adopt it and what users should know before letting an agent act for them.

Contents

What Poppy 1.0 is

The Poppy specification says it “defines how a Personal Agent, acting for one User, works with a Company on that User’s behalf.” It covers discovery, identity, account access and interactions across APIs, web browsing and conversations with the company’s own agent.

Sierra’s release post, by Bret Taylor and Clay Bavor, lists four launch facts:

  • Version 1.0 of the protocol is available.
  • Shopify has enabled Poppy for every merchant on its platform.
  • Poppy is on for everyone who uses Muse, Meta’s AI agent, which looks for Poppy first on every site it visits.
  • 60 new design partners joined, bringing the total to more than 100, including Delta Air Lines, Intuit, Salesforce, Twilio, Wayfair and Zoom.

Mercury, Synchrony and Turkish Airlines are among the partners going live. If you use Muse, see our Muse Spaces guide and Muse Mail guide for related features.

What changed in version 1.0

Sierra says version 1.0 reflects five weeks of design partners testing drafts against real websites, APIs and agents. Since the first draft, the protocol:

  • Lets companies adopt it in stages, starting with the basics.
  • Refines how it uses existing standards such as MCP, OpenAPI and UCP.
  • Gives companies more ways to guide the agent and collect feedback from it.
  • Adds extensions, such as a commerce extension co-authored by Shopify, so industries can build on the core without changing it.

The specification describes 1.0 as version 0.4 “with the fixes implementers needed.” It also sets a compatibility rule: from 1.0, a newer minor version will not break older personal agents, which ignore fields they do not recognize.

How the protocol works

The specification splits Poppy into layers that a company can choose from. They are not a fixed adoption sequence.

Component What it does
Discovery The company publishes /.well-known/poppy.json, listing its APIs, website, company agent and the access agents can request.
Operator recognition A recommended Web Bot Auth signature lets existing bot defenses recognize the agent’s operator.
Sessions A signed Poppy-Session header names the agent and a pseudonymous user ID that differs at each company.
Account access Standard OAuth with user-approved scopes; signed-out use needs no OAuth server.
Interaction channels OpenAPI or MCP APIs, website browsing, and conversations with the company agent, including handoff to a person.
Task coverage and extensions Lists tasks the company does not offer, collects feedback and adds features such as operation approvals.

The extensions page lists six 1.0 extensions: browser-control, disclosures, links, mediated-sign-in, operations (version 3) and ucp.

How a company adopts Poppy

Sierra says any company can now implement Poppy on its own, and its implementation kit includes a validator. Based on the specification and the official implementation guides, a typical path looks like this:

  1. Decide which tasks agents should handle and which interfaces you already have: APIs, an MCP server, a website or a company agent.
  2. Publish poppy.json over HTTPS at /.well-known/poppy.json with protocol_version 1.0, an organization block and at least one of agent, apis or web.
  3. To offer sign-in, publish OAuth protected resource metadata and authorization server metadata whose protected_resources lists your domain. The guides say this field is the one an existing issuer is most likely to be missing.
  4. Describe each scope in plain language and add scope_hints that mark whether it makes irreversible changes or exposes personal data.
  5. Verify the Poppy-Session header, ideally at one gateway, before relying on session identity or state.
  6. Optionally add a browser session endpoint, a tasks endpoint for excluded tasks and feedback, and extensions such as operations.
  7. Run the validator from Sierra’s implementation kit and fix what it reports.

What it means for users

For users, the main change is control. Instead of an agent being blocked or signing in with your full password access, you sign in on the company’s own page and approve specific scopes. The specification calls Direct Sign-In the preferred method.

Signing out revokes the agent’s refresh tokens, and the guides recommend that companies list each connected agent in account settings with a disconnect control. Under the operations extension, the user, or the agent following the user’s instruction, approves exact terms such as an amount.

Live examples from Sierra: Mercury customers can ask an agent to check balances or start a payment that still goes through the account’s approval policy. Turkish Airlines lets agents check fares and find business-class upgrades that customers approve before paying.

Poppy compared with MCP and UCP

Sierra describes UCP and Poppy as complementary: UCP handles the shopping itself, while Poppy handles what happens when an agent arrives, identifies itself and finds out what it may do. MCP is one of the two API types a poppy.json file can list, alongside OpenAPI. For other agent payment approaches, see our ChatGPT Wallet guide and Replit agentic commerce guide. For MCP apps inside a chat assistant, see our Claude Interactive Connectors guide.

Limitations

  • Adoption is early. Muse is the agent Sierra names as using Poppy for everyone, and only some design partners are live.
  • Draft label. The Next Web notes that the specification is still labeled a draft and allows changes that could break existing setups.
  • Transitional sign-in. The guides say Mediated Sign-In is planned for retirement on March 17, 2027.
  • Names are not verified. The guides warn that a valid poppy.json proves a domain publishes it, not that the domain belongs to the named company.

Frequently asked questions

What is the Personal Agent Protocol (Poppy)?

Poppy is an open protocol that defines how a personal AI agent, acting for one user, discovers a company and works with it through APIs, the company's website and the company's own agent. Sierra released version 1.0 on November 13, 2026, and the specification is published under the Apache License 2.0.

Which AI agents support Poppy 1.0?

Sierra says Poppy is now on for everyone who uses Muse, Meta's AI agent, and that Muse looks for Poppy first on every site it visits. Other personal agents can implement the protocol from the public specification.

Do Shopify merchants need to set up Poppy?

No. Sierra says every Shopify store now publishes a Poppy file automatically, so merchants do not need any extra implementation. Shopify also co-authored the commerce extension that links Poppy with the Universal Commerce Protocol.

Does my personal agent get my password?

Not with Direct or Device Sign-In. You sign in on the company's own page through standard OAuth and choose which scopes to grant. A transitional Mediated Sign-In extension lets an agent sign in with credentials you provide, and the guides say it is planned for retirement on March 17, 2027.

How is Poppy different from MCP?

MCP is one of the API types a company can list in its poppy.json file. Poppy sits around it, covering discovery, user sign-in, sessions that span APIs, web browsing and conversations, and rules for what the agent may do.

Does Poppy cost anything?

Sierra has not announced any fee for using the protocol. The specification and documentation are free to use under the Apache License 2.0, although the Poppy name and logo are not covered by that license.

Sources

Related answers

Leave a Comment