A2A Protocol Reaches v1.0 — The Open Standard for AI Agent Communication Now Has 150+ Backers

The Agent-to-Agent (A2A) Protocol hit v1.0 on April 9 — its one-year anniversary — and the release matters for anyone building multi-agent AI systems. This is the open standard for letting AI agents from different vendors, frameworks, and organizations communicate with each other across organizational boundaries. With 150+ supporting organizations signed on (AWS, Cisco, Google, IBM, Microsoft, Salesforce, SAP, ServiceNow), A2A is becoming the HTTP of the agentic web.

What Changed in v1.0

The stable release introduces four major additions over the earlier 0.x specifications.

Multi-protocol support lets A2A agents communicate over HTTP, WebSocket, or gRPC depending on the use case — HTTP for request-response, WebSocket for streaming, gRPC for low-latency internal coordination. Enterprise multi-tenancy was a recurring request from large deployments: organizations can now isolate agent communication across business units without running separate infrastructure. Modernized security flows replace ad hoc token handling with a standardized OAuth2 and MTLS path. And critically, there is now a formal migration guide from pre-1.0 implementations — which matters because A2A already has production deployments at several founding organizations.

Production SDKs are available in Python, JavaScript, Java, Go, and .NET as of today.

How A2A Differs From MCP

The comparison that keeps coming up is between A2A and the Model Context Protocol (MCP), which recently crossed 97 million monthly npm downloads. They are complementary, not competing.

MCP solves the agent-to-tool problem: it gives AI models a standard interface for connecting to external data sources, APIs, and services. A2A solves the agent-to-agent problem: it defines how two separate AI agents coordinate tasks, delegate subtasks, and pass structured outputs across organizational and vendor boundaries.

The practical use case: an enterprise orchestrator agent (built on LangChain, built internally, whatever) dispatches a subtask to a specialized coding agent from Cursor or a data retrieval agent from Salesforce. A2A is the protocol that makes that handoff well-defined, auditable, and vendor-agnostic. Without it, every cross-agent integration is a custom bespoke bridge — exactly the same mess HTTP replaced for web APIs in the 1990s.

Why 22,000 GitHub Stars in One Year Matters

A2A’s adoption curve mirrors what happened to MCP six months after its launch. Both protocols solve a real coordination problem that emerged as multi-agent systems moved from demos to production. The difference is that A2A started with enterprise buy-in from day one — Google, Microsoft, and IBM were founding members — whereas MCP grew bottom-up from developer adoption before large organizations followed.

The practical implication: if you are designing a multi-agent architecture today, evaluate A2A as your inter-agent communication layer. The protocol’s institutional backing means it will not disappear, and v1.0 means the spec is stable enough to build on without breaking changes. The 0.x-to-1.0 migration guide exists for a reason — there are already production systems to upgrade, not just demos.