Agentic Resource Discovery: Standardizing AI Capability Sharing

Key points

• 1

  What ARD solves: Before ARD, agents struggled to find and trust capabilities across organizations because each platform had its own internal registry. Now, ARD provides a standardized way for agents to discover capabilities by publishing them under an organization's domain name, which acts as the cryptographic foundation for trust. This means an operations agent investigating a production issue can dynamically find observability tools, documentation, or troubleshooting agents without being limited to one ecosystem. For example, an agent could query a registry to find a HIPAA-compliant healthcare tool from a different organization and securely connect to it using the tool's native protocol, eliminating the need for manual configuration or trust checks.

• 2

  How ARD works: ARD uses two key components: catalogs and registries. Catalogs are published by organizations under their own domain (e.g., `example.com/ai-catalog.json`), describing available capabilities like MCP servers or OpenAPI tools. Registries crawl these catalogs and index them for search, returning capabilities with verifiable trust metadata. When an agent needs a capability, it can either search registries using plain-language intent or directly fetch a catalog from a known partner's domain. Crucially, ARD includes cryptographic verification to confirm the publisher's identity before connecting, ensuring security without requiring the agent to handle complex authentication. This allows enterprises to onboard capabilities into a federated network while maintaining compliance, such as using Agent Registry to enforce HIPAA standards for healthcare tools.

• 3

  How Google Cloud implements ARD: Google Cloud's Agent Registry in Gemini Enterprise Agent Platform provides hosted support for ARD, enabling enterprises to publish, discover, and manage capabilities at scale. It assigns globally unique namespaced URNs, enforces egress policies, and pins tools to specifications, ensuring compliance with standards like HIPAA. For instance, an enterprise can onboard a healthcare agent to Agent Registry, which then verifies its trust manifest using cryptographic identity checks. This allows organizations to securely connect internal capabilities to the broader ARD network without rebuilding their own infrastructure. Native ARD support will roll out soon, letting businesses integrate their registries with the global network while maintaining control over governance and security.

What changed