Free AI check Book a strategy call

Agentic resource discovery checker

The agentic resource discovery checker tests whether an AI agent could find a machine-readable catalog of what your site offers. It requests /.well-known/ai-catalog.json and checks it is valid JSON, then looks for three pointers to it: a link in the homepage HTML, a mention in robots.txt and an HTTP header. It shows the catalog when found.

Signals
4 discovery signals
Convention
emerging, not ratified
Cost
free, no signup

Built and maintained by Sakshi Kumari and Ashish Khan, SEO specialists at MoxSEO. Last reviewed 19 October 2026.

The checker tests the origin of whatever URL you enter, so a domain is enough.

Try

Four discovery signals

Open any signal for why it matters and how to add it

  1. Not runCatalog at /.well-known/ai-catalog.json—

    Why it mattersA file at a predictable path is the simplest way for an agent to find what a site declares, without crawling it first. The checker needs an HTTP 200 response that parses as JSON.

    How to fixPublish a JSON file at /.well-known/ai-catalog.json listing only the public resources you want agents to use, and serve it with a 200 status and no login or challenge page.

  2. Not runHTML link to the catalog—

    Why it mattersAn agent that lands on your homepage can follow a link element in the head, the same way browsers find feeds and manifests.

    How to fixAdd a link element to the homepage head with rel="ai-catalog" (the checker also accepts rel="agentic-resource") and an href to /.well-known/ai-catalog.json.

  3. Not runrobots.txt mentions the catalog—

    Why it mattersCrawlers read robots.txt first, so a mention there is another place a client may notice the catalog.

    How to fixAdd a comment or line in robots.txt that names the catalog URL. Make sure robots.txt does not disallow /.well-known/ for the agents you want to reach it.

  4. Not runHTTP header points to the catalog—

    Why it mattersA response header works for every page and every client that reads headers, without parsing HTML.

    How to fixSend a Link header such as Link: </.well-known/ai-catalog.json>; rel="ai-catalog" from your server or CDN.

Catalog contents

The JSON served at /.well-known/ai-catalog.json, when found

Nothing to show yet. After a check, the catalog JSON appears here if the site serves one, so you can review exactly what it declares.

What is agentic resource discovery?

AI agents increasingly act on a user's behalf: searching, booking, calling tools. To do that they need a reliable way to find what a website, application or public service can do. Agentic resource discovery is the idea of publishing that as a machine-readable catalog at a predictable address, and pointing to it from the places agents already look.

The ai-catalog.json file at /.well-known/ is an emerging convention, not a ratified standard. Adoption and client support are limited, and no major assistant is known to rely on it today. Checking for it costs nothing; building it is worthwhile mainly if you already offer public tools, APIs or datasets you want agents to use.

For how agents find APIs and resources, read our guide to agentic resource discovery.

The four signals the checker looks for
SignalWhereFound when
Catalog file/.well-known/ai-catalog.jsonThe request returns HTTP 200 and the body is valid JSON.
HTML linkHomepage HTMLA link element has rel ai-catalog or agentic-resource, or an href to ai-catalog.json.
robots.txt/robots.txtThe file mentions ai-catalog.
HTTP headerHomepage response headersAn ai-catalog header, or a Link header mentioning ai-catalog.

How to use this result

  1. Step 1

    Start with the canonical origin

    Test the public domain that should own the capability declarations, and confirm redirects keep /.well-known/ reachable on the final host.

  2. Step 2

    Distinguish declaration from availability

    Finding a catalog shows that something is advertised. It does not prove authentication, schema quality, tool safety or that anything works when called.

  3. Step 3

    Review the catalog contents

    Read the JSON in the details panel. Every resource it lists should be public, accurate, owned by you and current.

  4. Step 4

    Validate every declared endpoint

    Check status codes, content types, permissions, descriptions, input expectations, ownership, versioning and failure behaviour for each resource.

  5. Step 5

    Add the pointers

    Once the file is right, add the HTML link, the robots.txt mention and the Link header, then re-run the check.

Discovery is not a security review

Do not expose private tools, credentials, administrative endpoints or sensitive capability descriptions merely to improve discovery. Public machine-readable resources need the same security and access review as any other production surface.

If agents need your site's knowledge rather than executable tools, a compact Markdown bundle that keeps source URLs may serve better. Build one with the OKF Bundle Generator, and make sure discovery files are reachable under your crawler rules with the AI Crawler Access Checker.

What the checker does

  • Requests /.well-known/ai-catalog.json and checks for 200 and valid JSON
  • Reads the homepage HTML for a catalog link
  • Reads robots.txt for a catalog mention
  • Reads response headers for a catalog pointer
  • Shows the catalog JSON when found

What it does not do

  • Validate the catalog against a schema, because no ratified one exists
  • Call or test the resources the catalog lists
  • Check authentication or permissions
  • Tell you whether any assistant reads the file

Need help reviewing discovery signals?

Review the published discovery resources and access signals in the context of your website and intended agents.

Discuss this with us

Agent discovery, answered plainly

Short, standalone answers. Email editorial@moxseo.com if any of them is wrong.

What does the Agentic Resource Discovery Checker check?

It checks four signals on a site's origin: whether /.well-known/ai-catalog.json returns HTTP 200 with valid JSON, whether the homepage HTML links to the catalog, whether robots.txt mentions it, and whether an HTTP header or Link header points to it. It shows the catalog contents when one is found.

What is ai-catalog.json?

It is a JSON file served at /.well-known/ai-catalog.json that lists the resources, tools or data a site makes available to AI agents. It is an emerging convention rather than a ratified standard, so its format and client support are still settling.

Do ChatGPT or Claude read ai-catalog.json?

There is no public confirmation that major assistants rely on it today. Adoption is limited. Treat the file as low-cost preparation if you offer public tools or data, not as a way to gain visibility now.

Is a missing catalog a problem?

Not for most sites. If you do not offer public tools, APIs or datasets for agents to use, there is nothing to declare. For content, a clear site structure, llms.txt and accessible pages matter more.

Does finding a catalog mean the tools work?

No. The checker only confirms that a catalog is advertised and parses as JSON. It does not call any listed resource or check authentication, schema quality, safety or whether anything works when an agent uses it.

What should never go in a public catalog?

Private tools, credentials, administrative endpoints and sensitive capability descriptions. A public catalog is a production surface and needs the same security and access review as any public API.

All 24 free tools

Planning for AI agents?
Start with what is real.

A free 30-minute call with Sakshi Kumari or Ashish Khan. They will separate the agent signals worth building now from the ones still settling, and check your site is reachable and readable first.

Book a strategy call AI SEO services
A call is not an engagement. We never guarantee rankings or AI citations.