Google Cloud has announced the release of Agent2Agent (A2A), an open protocol that facilitates communication among AI agents.

This protocol was developed with support from over 50 technology partners, including Atlassian, PayPal, Salesforce, SAP, and Workday, as well as consulting firms like Accenture, Deloitte, and Infosys.

A2A allows AI agents from different vendors to securely exchange information and coordinate actions across enterprise platforms.

Built on established standards such as HTTP and JSON-RPC, it is designed to integrate with existing IT systems and includes enterprise-grade security features.

The protocol also supports long-running tasks, providing real-time updates and feedback.

Core features of A2A include task management, capability discovery, and collaboration between agents. Agents can share their capabilities using a JSON-formatted “Agent Card,” which helps identify the most suitable agent for specific tasks.

🔗 Source: Google

🧠 Food for thought

1️⃣ Standardization efforts follow familiar patterns across technologies

Google’s Agent2Agent protocol mirrors interoperability challenges in healthcare, where data silos similarly limited innovation potential for decades.

Healthcare’s journey toward standards like FHIR (Fast Healthcare Interoperability Resources) since the mid-2010s demonstrates both the necessity and difficulty of creating common protocols across complex systems 1.

The emphasis on HTTP, SSE, and JSON-RPC in A2A’s design reflects lessons from previous standardization efforts, building on existing technologies rather than creating entirely new frameworks 2.

This pattern of standardization typically emerges when technology ecosystems reach a maturity point where fragmentation begins limiting further advancement, as seen with web standards and mobile OS interfaces.

Similar to healthcare’s experience, the success of A2A will likely depend on achieving both syntactic interoperability (common data formats) and semantic interoperability (shared meaning of data), particularly challenging with AI’s contextual understanding 2.

2️⃣ AI agent ecosystem follows platform industry patterns

The competitive landscape for AI agents has rapidly fragmented, with multiple companies developing proprietary frameworks rather than unified standards, creating market inefficiencies 3.

Recent Google developments