A free GeoSuite tool

Check and create your poppy.json for the Personal Agent Protocol

A person’s own AI agent reads /.well-known/poppy.json to find a company, sign in for that person over OAuth and act for them. Enter your site: we check the file the way an agent would.

Free
No sign-up
Open source (MIT)

The Personal Agent Protocol by Sierra and Meta is a draft (0.1, 9 October 2026). Poppy Check is an independent GeoSuite tool, not affiliated with either.

What an agent reads/.well-known/poppy.json

{
  "protocol_version": "0.1",
  "organization": {
    "name": "Acme",
    "domain": "acme.com"
  },
  "auth": {
    "issuer": "https://auth.acme.com",
    "direct": {
      "scopes": ["poppy:read", "poppy:write"]
    }
  },
  "apis": [
    {
      "type": "mcp",
      "url": "https://acme.com/mcp",
      "description": "Orders and returns",
      "resource": "https://acme.com/mcp"
    }
  ]
}

Create your poppy.json

Fill in what you offer and the page builds the file in your browser: nothing you type is sent anywhere. Then publish it at /.well-known/poppy.json and add two fields to your sign-in server’s OAuth metadata.

Run it from your terminal

The same checker and generator as a zero-dependency CLI on npm: check after a deploy, validate before you publish, init to start a file.

npx @geosuite/poppy-json check yoursite.com

npx @geosuite/poppy-json validate poppy.json --domain yoursite.com

npx @geosuite/poppy-json init --name "Acme" --domain acme.com --issuer https://auth.acme.com --mcp https://acme.com/mcp

Star on GitHub npm

Agents can find you. Do AI assistants recommend you?

A poppy.json lets a customer’s own agent act for them. First, someone has to choose you: GeoSuite asks ChatGPT, Gemini and Perplexity the questions your customers really ask, and tells you whether you are recommended, who wins instead and what to fix first. The report arrives by email.

Get the free analysis