The most important framework work often looks boring from the outside. A2A reaching a stable, cross-vendor, Linux Foundation-governed phase is one of those developments. Zero-human companies will eventually depend on agent interop the way cloud software depends on APIs.
What Changed
On April 16, 2026, Google's open source team published a one-year update on the Agent2Agent protocol. Google says the project now sits under Linux Foundation stewardship, has support from over 100 technology companies, and reached a stable v1.0 release in March.
Google also highlights that the founding coalition includes AWS, Cisco, Microsoft, Salesforce, SAP, and ServiceNow. That is not a side-project ecosystem. That is the sort of vendor spread you need before a protocol can credibly become infrastructure.
Why This Matters
The typical agent demo still assumes a single model talking to a handful of tools. Real autonomous companies will not work that way. They will use specialist agents for sales, finance, research, support, compliance, and engineering, all passing work back and forth across boundaries.
That means we need standards for discovery, delegation, trust, and structured exchange between agents that were not built by the same vendor and do not live in the same runtime. A2A is one of the clearest attempts to define that layer.
How It Fits with IZHC Coverage
This update sharpens themes we already covered in Microsoft's governance toolkit and in OpenAI's agent infrastructure shift. Governance answers how agents should be controlled. Managed runtimes answer how they should execute. A2A starts answering how they should coordinate.
Google's update also points to an expanding protocol family including A2UI, AP2, and UCP. Whether those specific variants win or not, the direction is clear: the market is trying to standardize the seams between autonomous systems.
The Take
A stable agent protocol is not exciting in the same way a new frontier model is exciting. But it may prove more durable. Shared standards lower switching costs, reduce vendor lock-in, and make it easier for specialized agent companies to plug into each other.
That is exactly the kind of framework maturity zero-human companies need. If the future is multi-agent, the protocol layer is not optional.
Related: See our earlier research on Microsoft's governance toolkit, Responses API, and the April 2026 briefing.