Agent Registry data model

Agent Registry categorizes agentic components of your ecosystem into four main resource types that you can manage and discover:

• Agents: Autonomous AI actors that carry out specific tasks, invoke tools, and collaborate with other agents.
• MCP servers: Providers of standardized data resources and tools. Tools are deterministic functions exposed by an MCP server that agents can invoke to perform specific actions.
• Endpoints: Target URLs, typically REST APIs, accessed by an agent. By abstracting these destinations into manageable resources, Agent Registry lets you centrally govern which external APIs your agents can access.
• Skills: Modular capabilities that extend what an agent can do. Agent Registry distinguishes between standalone skills, which are packages of instructions and optional scripts governed directly in the registry, and A2A skills, which are capability descriptions listed on an agent's Agent Card that tell other agents when and how to communicate with it.

For more information about these and other concepts that Agent Registry uses, see Key concepts.

API resources

The Agent Registry API separates read and write operations for agents, endpoints, and MCP servers:

• Writable API resource: You manage these agentic components by manually creating, updating, or deleting a Service resource in the registry.
• Read-only API resource: To query, list, or search for these agentic components, you request the read-only Agent, McpServer, or Endpoint resource to the Agent Registry API.

The specification flags you use when creating a Service determine the resulting read-only resource collection:

Flags	Resource collection	Use case	Valid values
--agent-spec-type --agent-spec-content	Agent (agents)	Remote agents (A2A-compliant or standard REST agents)	a2a-agent-card, no-spec
--endpoint-spec-type --endpoint-spec-content	Endpoint (endpoints)	External target REST APIs or service URLs	no-spec
--mcp-server-spec-type --mcp-server-spec-content	McpServer (mcp-servers)	External Model Context Protocol (MCP) servers	tool-spec

These flag pairs are mutually exclusive. If you specify invalid flags, your resource is created in the incorrect collection and might not work with type-specific policies.

Standalone skills don't use the Service projection model. Instead, you manage, version-control, and list skills directly using the Skill, SkillRevision, and Publisher API resources.

To get started, see Set up Agent Registry. For more information about the Agent Registry API, see the API reference documentation.

Supported products and components

Agent Registry categorizes supported products and components by how they are registered: automatically or manually.

Automatic registration

Agent Registry automatically discovers and ingests supported Google Cloud workloads deployed within your project. For agents that implement the open Agent2Agent (A2A) protocol, the registry automatically extracts their A2A skills and capabilities from their Agent Card. Google-created standalone skills are also globally discoverable when you set up Agent Registry.

The following table lists the products and components that Agent Registry registers automatically:

Component type	Supported products
Agents	Agent Runtime on Gemini Enterprise Agent Platform Google Workspace Gemini Enterprise Google Kubernetes Engine (GKE) deployments including a specific functional type label Cloud Run workloads configured with a specific functional type flag
MCP servers	Official Google and Google Cloud remote MCP servers Google Kubernetes Engine deployments including a specific functional type label Cloud Run services configured with a specific functional type flag
Skills	Google's public skills are globally discoverable from all projects

Manual registration

For custom workloads, external cloud runtimes, on-premises infrastructure, or agents that don't implement the A2A protocol, you manually register components using the writable Service resource. For custom standalone skills, you use the Skill resource.

The following table lists the products and components that you register manually:

Component type	Supported resources
Agents	Remote agents deployed on Compute Engine, Google Kubernetes Engine, Cloud Run, on-premises systems, or external clouds
MCP servers	Custom agentic components on unsupported runtimes or external MCP servers
Endpoints	External target REST APIs or service URLs governed as endpoints
Skills	Your own standalone skills authored by your organization

What's next

• Review Key concepts to familiarize yourself with the essential terms used in Agent Registry.
• Learn about Identity and identifiers.
• Set up Agent Registry.
• Learn about the difference between automatic and manual registration.