Glossary
A2A Protocol
An emerging open standard for how autonomous AI agents advertise their capabilities, discover one another and exchange work, complementing tool-connection protocols such as MCP.
General definition
The A2A protocol, short for agent-to-agent protocol, is an emerging open standard for communication between autonomous AI agents. Where the Model Context Protocol (MCP) connects an agent to tools and data, A2A connects agents to each other as active parties that can reason, hold state and initiate their own requests.
- An agent card: a self-published description of an agent’s skills, input and output formats and authentication
- Discovery, so an agent can find another with the right capability
- A request and response flow for delegating work and returning results
- A foundation to build on, rather than a complete answer to coordination
A2A is still early. It gives agents a way to describe themselves and exchange messages, but the card is self-advertised, so trust has to come from somewhere else. A durable system treats observed capabilities as more reliable than advertised ones, and layers reputation and agent social memory on top, so that an agent remembers who was actually reliable rather than who claimed to be.
In the Ethora ecosystem
Ethora treats AI agents as first-class participants in the same real-time messaging fabric as people. An agent built with the AI SDK joins chat rooms and talks to humans and to other agents over the same XMPP-based channel, with presence, history, delivery and an audit log handled for you. That gives multi-agent coordination the messaging fundamentals a bare A2A transport does not, and a natural substrate for the reputation and memory layers that make agent-to-agent work trustworthy.
