Google's Agent2Agent protocol, the open standard for cross-vendor agent communication, has moved past the pilot stage into real production use. According to reporting on Google Cloud Next 2026, the protocol has reached 150 organizations in production, not pilot, routing real tasks between agents built on different platforms, and is now governed by the Linux Foundation's Agentic AI Foundation. The protocol itself has advanced to version 1.2, with signed agent cards using cryptographic signatures for domain verification. That last detail matters more than the adoption numbers. An A2A "agent card" is the discovery document an agent publishes describing its capabilities, endpoints, and how to reach it — the thing another agent reads before deciding to hand off a task. Up to this point, nothing stopped an attacker from publishing a convincing-looking card claiming to represent, say, a bank's customer-service agent. Cryptographic signatures tied to domain verification close that gap the same way TLS certificates tied HTTPS sites to registered domains: a receiving agent can now check that the card actually originates from the domain it claims, before trusting anything in it. The production names are notable too. Microsoft, AWS, Salesforce, SAP, and ServiceNow are running A2A in production environments, which means this isn't a demo-only feature — it's shipping in systems that route actual enterprise workflows. The framing from Google Cloud's keynote was explicit about positioning: A2A is designed to complement rather than compete with Anthropic's Model Context Protocol, with MCP handling agent-to-tool connections and A2A handling agent-to-agent handoffs. What this signals for the broader agent ecosystem: discovery and identity are converging at the protocol layer rather than being bolted on separately. A card that can be cryptographically tied to a verified domain is effectively a lightweight trust anchor — closer to how a CA-issued certificate lets a browser trust a website than to today's ad hoc API-key trust models between agent frameworks. It's a small mechanical change but it's the kind of change that has to happen before agent-to-agent commerce or delegated task execution can scale past controlled pilots, because right now there's no way for an agent to know it's actually talking to the counterparty it thinks it is. Signed, domain-verified cards are a first concrete answer to that problem at protocol scale, rather than something each vendor solves independently.