Dynamic Feed

Google just published Agentic Resource Discovery — an open standard (with the Linux Foundation) for how AI agents find and connect to capabilities across the web. Think DNS + a phone book for agents.

The model is simple: you publish an ai-catalog.json at a well‑known path on your domain → registries crawl it → an agent searches by intent → it verifies the publisher → then it connects (over MCP, A2A, or a plain API). Publish → crawl → search → verify → connect.

Most of the coverage will fixate on discovery. I want to flag step 4, because it's the one that actually matters: verify. The agent is about to act on what it finds — call a tool, pull data, make a decision. "I found a capability" is worthless if it can't check who published it and whether it's been tampered with. Discovery without verification just industrialises trusting strangers.

That's the half we've been building the whole time — so when the spec dropped this week, we shipped our catalog the same week. Dynamic Feed is now a verifiable ARD publisher: domain‑anchored identity, every datapoint Ed25519‑signed and checkable against our public key.

The agentic web is going to be discovered and verified. Good to see the standard treat the second half as a first‑class citizen.

If you publish an API, an MCP server, or an agent — you can host an ai-catalog.json today. It's a single file. Spec's open: github.com/ards-project/ard-spec