Overview
- Google published the Agentic Resource Discovery (ARD) specification as an open Apache 2.0 standard that defines how organizations publish searchable catalogs of agentic resources using a machine-readable ai-catalog.json file hosted on their domains.
- ARD uses publisher-hosted catalogs as the cryptographic root of trust and separate registries to crawl, index, and return verifiable metadata so an agent can decide which external tool to use before connecting.
- Google Cloud already backs the model with its Agent Registry inside the Gemini Enterprise Agent Platform, which offers hosted search, globally namespaced URNs, egress policy controls, pinned artifacts, and Agent Identity verification for compliance.
- Planned product updates include authenticated publisher onboarding and native ARD support integrated into the Agent Platform in the coming months to extend federated discovery and enterprise verification workflows.
- ARD builds on the AI Catalog data model from the AI Catalog Working Group under the Linux Foundation and publishes quickstart guides and a GitHub repo to encourage community federation, cross-provider discovery, and broader developer adoption.