Google launches the new Agent2Agent protocol enabling AI agents to communicate with each other, safely exchange information, and perform actions.
Google launches an open protocol, Agent2Agent (A2A), allowing AI agents to communicate and collaborate with each other. The A2A protocol is designed to improve the interoperability of autonomous AI systems within business environments. According to Google Cloud, the protocol enables AI agents to safely exchange information, coordinate tasks, and collaborate on workflows, regardless of the underlying framework in which they are built.
AI interaction
AI agents are increasingly being used to automate business processes. They can perform tasks such as managing orders, supporting customer service representatives, or improving supply chain processes. In practice, these agents often operate within separate systems or applications, which makes collaboration difficult.
With the A2A protocol, Google Cloud aims to remove these barriers. The protocol uses existing standards such as HTTP, SSE, and JSON-RPC and provides support for enterprise-level authentication. It is modality-agnostic and designed to support both quick and long-duration tasks.
read also
Google Ends Cloud Exclusivity for Gemini and Agentspace
“We believe that the A2A framework will add significant value for customers, whose AI agents will now be able to operate across their entire business application domain,” Google states in the announcement.
Google collaborated with more than 50 technology and service partners on the protocol, including Atlassian, SAP, Salesforce, PayPal, Box, Workday, Accenture, and Deloitte. The protocol is offered as open source and is expected to be production-ready later this year.
Task distribution
In practice, A2A enables a clear division of tasks between ‘client’ and ‘remote’ agents. A client agent formulates a task, while a remote agent executes it. Through the protocol, agents exchange information about their capabilities, progress, and the results of completed tasks. These results are recorded within the protocol as ‘artifacts’.
An example of an application is the recruitment process for software engineers. An AI agent can collaborate via A2A with other specialized agents to search for suitable candidates, schedule interviews, and coordinate background checks, all through a single interface.