Data

Who has published an ARD manifest

Adoption measured the only way that means anything: domains that actually serve a manifest. It is a smaller number than anyone quotes, and most of it is on the wrong path.

206 domains serve an ARD manifest at a well-known path, out of 375,997 crawled. That is the real adoption figure; larger publisher counts measure domains that appear in an index, most of which never published anything.

Of those 206, 176 (85%) are still on /.well-known/ai-catalog.json, the path v0.91 superseded on 26 August 2026, and only 29 have moved to ard.json. Consulting the old path is optional for consumers, so a fully conformant registry will not see any of them.

The honest adoption number

206domains serving a manifest

375,997domains crawled

176on the predecessor path

29on ard.json

Two months after the announcement, ARD adoption measured by the only thing that counts as adoption, a manifest served at a well-known path, is 206 domains. Everything larger you will read, including the publisher counts on this site's own data page, is measuring something else and says so.

Why this number is so much smaller than the other one. A publisher is any domain with an entry in an index. Most got there because somebody crawled the MCP Registry or an OpenAPI corpus, and never published a manifest at all. A manifest publisher put a file on their own domain on purpose. The first number is in the thousands, the second is in the hundreds, and only the second one means anyone adopted anything.

The migration gap

This is the finding worth carrying away, and it is measurable rather than asserted.

Path	Publishers	Share	Status
`/.well-known/ai-catalog.json`	176	85%	Predecessor. Superseded by v0.91 on 26 August 2026.
`/.well-known/ard.json`	29	14%	Current. What a conformant consumer is required to fetch.

85% of everyone who has published an ARD manifest is on a path the specification replaced. That includes GitHub's own MCP catalogue, which is the single largest manifest in the index.

None of them are doing anything wrong. The rename is nine days old at the time of writing, and a consumer may still consult the predecessor path, so nothing has broken yet. The exposure is precise and worth stating exactly: consulting it is optional. A registry that implements only the current specification, which is fully conformant, will not see any of those 176 publishers. The longer the gap persists, the more likely one of those registries is the one that matters to you.

Move to ard.json in ten minutes Check which path you serve

Who has published

The largest manifests in the index, by the number of resources each one describes. Ranked by what they publish rather than by what an index happens to hold about them: a cloud provider can have hundreds of entries catalogued from an OpenAPI corpus and no manifest of its own.

Domain	Resources	Path
`api.mcp.github.com`	250	predecessor
`gstcranes.com`	37	predecessor
`connectors-skills.zapier.com`	30	predecessor
`built2winweb.com`	28	predecessor
`datasignalslab.com`	19	predecessor
`zapier.com`	14	predecessor
`ultimaluz.com`	14	predecessor
`apify.com`	13	predecessor
`clickhouse.com`	12	predecessor
`eprimo.de`	12	predecessor
`hostinger.com`	12	predecessor
`www.pathwren.workers.dev`	12	ard.json
`anahana.com`	10	predecessor
`jacobcampbellrealtor.com`	10	predecessor
`mandateshield.com`	10	predecessor
`airtable.com`	10	predecessor

Measured 4 September 2026. The full list, with a page per publisher, is at neuronto.com/ard-publishers.

The shape of it

Most manifests are small, which is what you would expect this early: somebody publishing the one thing they built, not an organisation cataloguing a portfolio.

Resources in the manifest	Publishers
Exactly one	42
Two to four	82
Five to nineteen	78
Twenty or more	4

The long tail is the interesting part. 124 of 206 publishers describe four resources or fewer, and a manifest with one entry takes about ten minutes to write. Whatever else is true about ARD, the cost of joining is not what is holding adoption back.

Getting counted

You appear here when a crawler reaches your domain and finds a manifest. Publishing does not announce you, and no registry is watching your DNS.

Get indexed

You are in this count once a registry has crawled you, and a registry crawls domains it has encountered. One request is the difference.

one call, any ARD registry that accepts submissions

curl -X POST https://neuronto.com/submit \
  -H "Content-Type: application/json" \
  -d '{"url": "https://yourdomain.com"}'

Disclosure: Neuronto is run by the people who write this site, and it is named first for that reason as much as any other. It resolves a bare domain to your manifest, indexes entries of every type rather than MCP servers only, and answers with what it found or exactly what it tried. Add "dry_run": true to see the outcome without being written anywhere.

The other public registries take submissions through their own doors, and submitting to several costs nothing: WellKnown, ARD Registry Hub, Desvela. What actually reaches all of them over time is the manifest on your own domain, which is the part nobody can take away from you. How each one behaves, probed.

Method

Source. The crawl log of the public index operated by Neuronto, which is run by the people who write this site. That conflict is stated on the about page and on every page here that carries a measurement.
What counts. A domain is counted once if a fetch of a well-known manifest path returned 200 with a parseable JSON document containing an entries array. A domain that serves both paths is recorded at the one that answered first, which is ard.json.
What this misses. Domains nobody has crawled. The denominator is 375,997 domains checked, not the web, and a publisher no crawler has reached is invisible to this count and to every registry. That is the same problem the page is about.
Entry counts are what the manifest declared at crawl time, not what is live now. Liveness is on State of ARD.
Raw numbers: /api/stats.json, under manifests.

Questions

How many organisations have adopted ARD?

206 domains served a manifest at a well-known path when we last crawled, out of 375,997 domains checked. That is the honest adoption number. Counts of thousands of publishers, including on this site's other pages, measure something different: any domain with an entry in an index, most of which were catalogued from the MCP Registry or an OpenAPI corpus without ever publishing anything themselves.

Why are most of them on the old path?

Because the path changed on 26 August 2026 and crawlers still read the old one. 176 of 206 publishers are on /.well-known/ai-catalog.json. Nothing is broken for them yet; the risk is that consulting the predecessor path is optional for consumers, so a registry that only implements the current spec will not see them at all.

Is my domain in this count?

Only if a crawler has reached it. Publishing does not announce you. Check what your domain actually serves, then tell a registry.

Can I see the full list?

The full publisher index lives at neuronto.com/ard-publishers, with a page per publisher. This page is the analysis; that is the directory.

Last reviewed 2026-09-04. Checked against ARD v0.91 (Proposal, 2026-08-26).