Google Cloud Next 26: A2A Protocol and Gemini Enterprise Make AI Agents Interoperable
May 3, 2026
At Google Cloud Next 26 in mid-April 2026, Google introduced the Gemini Enterprise Agent Platform and the Agent2Agent protocol with 150 organizations in production. The push standardizes how AI agents talk to each other.
Google Cloud Next 26: Gemini Enterprise Agent Platform and Agent2Agent Protocol Version 1.2
At Cloud Next 26 in mid-April 2026, Google announced two building blocks that together reshape enterprise AI architecture. First, Vertex AI was renamed to the Gemini Enterprise Agent Platform and merged with the former Agentspace into a single product. Second, Google released the Agent2Agent protocol, A2A, in version 1.2. The protocol is now governed by the Linux Foundation under a dedicated Agentic AI Foundation, and according to Google, runs in 150 organizations in production rather than in pilot projects.
A2A 1.2 With Signed Agent Cards for Cross-Vendor Interoperability
A2A defines how agents from different vendors can talk to each other. That includes signed agent cards, meaning cryptographically verifiable identities. Microsoft, AWS, Salesforce, SAP, and ServiceNow already run A2A in production environments. The picture is clear: instead of a separate agent island per vendor, an open mesh is forming.
Workspace Studio, Project Mariner, and Apigee as Agent Tools
Inside the platform, Google introduces new tools. Workspace Studio is a no-code agent builder that turns typical business processes into reusable skills. Project Mariner is Google's web-browsing agent. With Apigee acting as an API-to-agent bridge and managed MCP servers, existing interfaces can be wired up without major rebuilds. According to Google, the Model Garden lists more than 200 models, including Anthropic's Claude.
Google's Competitive Strategy Against OpenAI and Anthropic
The strategic message is clear. While OpenAI and Anthropic emphasize their own models and tight integrations, Google is betting on a full stack with open interfaces that spans Microsoft to Salesforce. If that pays off, interoperability could become the decisive competitive lever.
Why it matters
For IT leaders, this means agent strategies should plan for multi-vendor scenarios. Choosing a single vendor today does not exempt later workflows from running over A2A. The security layer remains critical. Identities, permissions, and audit logs are mandatory, especially when agents from multiple vendors exchange data. Linux Foundation governance also creates a form of standardization that European and Asian providers will eventually align with.
Practical example
An international logistics company runs sales on Salesforce, HR on Workday, and warehouse management on SAP. Until now, these systems exchanged data through classic APIs, which required long integration projects. With A2A 1.2, a Salesforce sales agent can directly ask a SAP inventory agent whether a product is available, while simultaneously triggering a Workday agent to find a warehouse manager slot. Each agent verifies the request through a signed agent card. What used to take weeks of integration can now be set up in days, provided all three platforms are A2A-certified.
💡 In plain English
Imagine different helpers from different companies, each speaking a different language. Google built a translator language so that all the helpers can finish a project together. Now 150 big companies already do it this way.
Key Takeaways
- →Google Cloud Next 26 ran in April 2026 and unveiled the Gemini Enterprise Agent Platform.
- →Vertex AI is now Gemini Enterprise, with Agentspace folded in.
- →The Agent2Agent protocol reached version 1.2 and is in production at 150 organizations.
- →Microsoft, AWS, Salesforce, SAP, and ServiceNow run A2A in production.
- →Workspace Studio offers a no-code builder for business agents.