Overview
Extensions in the Agent-to-Agent Protocol allow agents to declare additional capabilities. They are intended to provide a client with metadata about which capabilities are supported, which are required, and documentation about the extension itself.Aion Components
Each component you see in the Project Editor (Architecture tab) is an A2A-compatible agent. This lets you communicate with independent pieces of your infrastructure as if your microservices were AI agents. A common pattern in Aion is that a component will directly communicate with one of its upstream components or to one located via a registry search in order to request additional context or services. Components define their own A2A Agent Card. These cards are used internally within the Aion control plane and may be composed into externally addressable Agent Cards when connected to a Distribution in the Project Editor (see Agent Sequences below). Each component can define its own supported or required extensions. This gives you a way to express protocol requirements between services, not just between a client and an agent.Note: By default, access to communicate directly with these components is scoped to other agents within the same project. We use each agent’s client ID and secret to authenticate. Each component deployed through our platform is automatically given these credentials. If you deploy a component remotely, we provide credentials at creation time.