With RemoteA2aAgent and A2A Protocol v1.0 now backed by 150+ organizations, agents written in different languages can discover, coordinate, and fail over without sharing a framework.
The bottleneck in modern agent architecture is no longer the model itself, but the rigid, hierarchical execution models. Google’s ADK 2.0, which reached general availability for Python on May 19, 2026, and for Go on June 30, 2026, addresses this by replacing traditional hierarchical executors with a graph-based engine. This engine supports sequential chains, conditional routers, fan-out/fan-in, nested sub-graphs, and first-class loops.
At the core is the RemoteA2aAgent abstraction. By automating the Agent Card handshake and managing parameter serialization and JSON-RPC 2.0 network requests, the SDK allows developers to treat any remote, A2A-compliant service as a local sub-agent. This decouples internal logic from the execution environment, enabling cross-language agent teams.
The A2A Protocol v1.0 has secured support from over 150 organizations, including AWS, Cisco, IBM, Microsoft, Salesforce, SAP, and ServiceNow. The ADK also natively supports MCP, granting access to over 17,000 MCP servers.
Rao Surapaneni of Google Cloud noted: “AI agents are only as useful as their ability to collaborate, and the adoption of A2A by more than 150 organizations underscores the widespread enthusiasm for an open, interoperable protocol.”