The ARD Standard

About this paper

Your agent can call anything. It cannot find anything. Every protocol in the agentic stack shipped its invocation layer and skipped discovery, which is why an agent’s world is still whatever a developer wired into it by hand. Agentic Resource Discovery is the first serious attempt to close that gap, and its answer is refreshingly cheap: publish a JSON file at a well-known path on your own domain, let anyone index it, and answer “what can help me do this?” over one endpoint.

Nine weeks after the draft landed, this report tells you who actually did it. Not who signed on — eleven of the largest agent builders in the world put a logo on the specification, and two of them serve a catalog on their own domains. The nine companies publishing one are mostly companies nobody in the working group has met, none of them is fully conformant, and two of the five failures were caused by the project’s own documentation rather than by the implementer. If you are deciding whether to publish a catalog, run a registry, or invest in the discovery layer, the useful question is not whether ARD is a good specification. It is whether anything reads it yet — and that answer is measured here, across 43,834 domains and four live registries.

What's inside

1. Executive summary — the ratio that runs backwards
2. What ARD is, and what it is not
3. The origin — the third time someone has tried to make the web describe itself
4. What the body actually publishes
5. The maturity ladder — there isn't one, and that is the process gap
6. Three places the documentation contradicts the tooling
7. The coalition — eleven logos, three commit bits
8. The adoption gap — nine publishers, and seven of them are outsiders
9. Conformance — a good tool, and what it cannot reach
10. The regulatory layer — nothing compels this, and one regime will touch it
11. The scoring caveat
12. What would have to be true
13. The investable thesis
14. Where this is going

What you get for $500.00