Google's Open Standard Is a Search Engine for AI Agents

Google has published the Agentic Resource Discovery (ARD) specification, an open standard that works like a search engine for AI agents — letting them locate, identify, and verify tools and capabilities across the open web rather than relying on hand-wired, one-off integrations. The spec is available now under an Apache 2.0 license and arrives backed by an unusually broad coalition of rivals.

Launch contributors include Microsoft, Nvidia, GitHub, Hugging Face, Cisco, Databricks, Salesforce, ServiceNow, Snowflake and GoDaddy. ARD builds on the AI Catalog data model from the AI Catalog Working Group under the Linux Foundation, a neutral-ground signal that Google wants this treated as shared infrastructure rather than a proprietary play.

Technically, ARD rests on two primitives. Organizations publish a machine-readable ai-catalog.json file at a well-known path on their own domain, describing the capabilities they expose — MCP servers, OpenAPI tools, agents, or nested catalogs. Registries then act as the search layer, crawling and indexing those catalogs so an agent can ask "where does the right capability live, which one should I use, and is it safe to connect to?" Domain ownership underpins a cryptographic identity layer, so an agent can verify a publisher before it ever opens a connection.

Crucially, ARD does not compete with existing agent plumbing like Google's own Agent2Agent (A2A) protocol or Anthropic's Model Context Protocol (MCP) — it sits a layer above them as the discovery and trust step, then hands off to those native protocols for the actual runtime connection. Google is already shipping it: its Cloud Agent Registry inside the Gemini Enterprise Agent Platform implements ARD with namespaced URNs, egress policies and cryptographic trust manifests, while Hugging Face has launched a Discover Tool that wraps semantic search over its Hub Spaces, Skills and MCP servers in the ARD envelope.

The bigger picture is what the industry has been missing as agents proliferate: a common way to find and trust capabilities across organizational boundaries. If ARD gains traction, it could become the connective tissue — a kind of DNS-plus-search layer — that lets autonomous agents safely discover one another's tools at internet scale, rather than every vendor building bespoke bridges. That so many direct competitors signed on at launch is the strongest sign yet that agent interoperability is becoming table stakes.