A2A passed 150 supporting organisations, and NIST started its own initiative
Google's Agent2Agent protocol went from about 50 supporters at launch to more than 150 by April, including AWS, Microsoft, Salesforce and SAP. NIST opened an AI Agent Standards Initiative in February.
Agent2Agent counted more than 150 supporting organisations by April, up from around 50 at its launch a year earlier. The membership list is the interesting part: AWS, Google, Microsoft, Salesforce and SAP are all in it, which is to say the major clouds and the major systems of record.
That combination matters because agent-to-agent communication is only useful if the agents belong to different companies. An agent talking to another agent inside the same vendor's product does not need a protocol; it needs a function call. The value appears precisely at the boundary between organisations, which is why the standard needed the competitors to sign.
Running alongside it, NIST opened an AI Agent Standards Initiative in February, aimed at interoperable and secure agent systems. A standards body entering while an industry consortium is mid-adoption usually produces one of two outcomes: the government work ratifies what shipped, or it splits the field.
The security half of the NIST framing is the part with no good industry answer yet. Interoperability has a business case that funds itself. Security across an agent boundary — establishing what a foreign agent is allowed to ask for, and proving who it acts for — costs money and delays launches.
Adoption counts also flatter themselves. "Supporting organisation" spans a company that has shipped A2A in production and one that put a logo on a page. The same caution applies to the standards announced this month.
NIST — Announcing the AI Agent Standards Initiative for Interoperable and Secure Innovation → · Google Developers Blog — Announcing the Agent2Agent Protocol (A2A) → · Rost Glukhov — Google A2A Protocol in 2026: Adoption, Hype, and Reality →