Agent Checked

Headless, custom and everything else

Is your custom or headless site agent-ready? Check it free

When you own the code, every pillar is yours. The check reads the same public documents an agent reads: robots.txt, the home page's JSON-LD, llms.txt, a sitemap, /.well-known/poppy.json and the issuer's metadata.

What the check usually finds on Headless, custom and everything else

Found
Whatever robots.txt says. Headless sites behind a CDN sometimes block unknown user agents at the edge, which the check reports as the home page not loading.
Understood
JSON-LD only if the front end renders it on the server; client-only rendering leaves the check (and agents) with an empty page.
Usable
No poppy.json unless you publish it. Serve the file yourself or proxy it from a hosted Company; a redirect to another host doesn't count.
Trusted
Nothing to score until there is agent sign-in.

What you can do

  1. 1

    Render structured data on the server

    Agents and the check read the HTML as served. Put Organization and Product/Service JSON-LD in the server-rendered page, not after hydration.

  2. 2

    Let assistants in through robots.txt

    Edit the file, and check any edge rule that blocks unknown user agents. Allow ChatGPT-User, Claude-User and Perplexity-User (they fetch a page only when a person asks) and the AI search bots (OAI-SearchBot, Claude-SearchBot, PerplexityBot). Blocking training crawlers such as GPTBot or CCBot is a separate choice and costs no points.

  3. 3

    Publish llms.txt

    A short Markdown page at /llms.txt naming the business, what it sells, its delivery and returns policies, and the pages that matter. Most platforms can serve a plain-text file at the root; where one cannot, a redirect to a page you host elsewhere works.

  4. 4

    The protocol part: a hosted Company

    One proxy rule for /.well-known/poppy.json at the edge, the web server or the framework's routes, so it is served on your own domain. Agent Adapt is building a hosted Company for sites that cannot run its WordPress plugin: it publishes poppy.json, runs the sign-in and the customer API, and the document is served on the site's own domain through a small proxy (the spec stopped allowing a redirect to the provider's host on 16 October 2026). It is not yet on sale.

    Ask for early access

Common questions

›Can I implement the Personal Agent Protocol myself?

Yes. The spec is open (Apache-2.0). It is an OAuth server with sender-bound (DPoP) tokens, agent client documents, a customer sign-in flow and a scope-enforced API. The check scores the discovery part; the rest you prove with a real agent.

›What does the check not test?

Anything that needs a session: it never signs in, never sends a token and never changes anything. It reads public documents only.

In depth: Headless and custom stores with personal agents on getagentadapt.com.

Other platforms