About Agent2Agent Protocol

This feature introduces a managed Agent2Agent (A2A) server in Autonomous AI Database that publishes Select AI agent teams as discoverable agents that clients can call.

External applications connect using REST APIs and OAuth authentication. Database credentials are used to obtain an access token, and all subsequent requests use the token instead of direct database connections. See A2A protocol to learn more.

A2A protocol is an HTTPS-based protocol that standardizes how AI agents discover, call, and coordinate across systems. Autonomous AI Database implements this protocol by exposing Select AI agent teams as A2A-compatible agents through managed REST endpoints.

The database supports defining, managing, and running agent teams that coordinate tasks, share context, and process multi-step workflows. When the A2A Server capability is enabled, A2A clients can discover available agent teams through Agent Cards, which describe capabilities, authentication requirements, and interaction patterns. The A2A server publishes these teams through agent discovery APIs and Agent Cards.

A2A client agents use OAuth-based authentication with database credentials to connect to the A2A Server and call agent teams. The interaction model supports asynchronous communication, enabling client agents to interact with the database agent teams through A2A server and retrieve results or status updates.

The service includes built-in governance using database roles, AI profiles, and auditing. Network controls such as private endpoints and access control lists restrict connectivity.

This A2A server removes the need for custom connectors and enables non-database applications to use Select AI capabilities through standard agent interfaces.