At the Cloud Next conference, Google presented a new open-source protocol Agent2Agent (A2A) aimed at ensuring safe interaction between artificial intelligence agents from different developers and technological frameworks. The protocol is based on HTTP, SSE, and JSON-RPC standards with authorization and authentication mechanisms.
Agent2A allows client agents to formulate tasks for remote agents and publish their capabilities in the form of agent cards (Agent Cards) described in JSON format. More than 50 partners have supported it, including Atlassian, Box, Cohere, Intuit, Langchain, MongoDB, Salesforce, SAP, ServiceNow, UKG, and Workday. Consulting companies such as Accenture, BCG, Capgemini, Cognizant, Deloitte, and PwC also joined the initiative.
Google published the complete A2A specification on GitHub, including code examples and use case scenarios. The company invites the open-source community to actively participate in developing the protocol. The production version is planned for release later this year.
Google believes that A2A will become a new industry standard for integrating artificial intelligence into corporate processes and everyday tasks.