Google Cloud Next 2026: AI agents, A2A protocol, Workspace Studio, and a comprehensive strategy to compete with OpenAI and Anthropic.

      Summary: At Cloud Next 2026, Google rebranded and unified its AI platform by changing Vertex AI to the Gemini Enterprise Agent Platform and integrating Agentspace into a single Gemini Enterprise product. Announcements included the no-code agent builder Workspace Studio, over 200 models in the Model Garden featuring Anthropic Claude, partner agents from Box, Workday, Salesforce, and ServiceNow, stable releases of ADK v1.0 in four languages, Project Mariner (a web-browsing agent), managed MCP servers with Apigee as a bridge for API-to-agent, and the A2A protocol v1.0 deployed across 150 organizations. Kurian emphasized a strategy that claims ownership of the entire stack from chip to inbox, contrasting with competitors who "give you the pieces, not the platform."

      During the opening keynote at Cloud Next 2026, Google presented a significant rebranding and consolidation of its AI platform centered around agents. Vertex AI has been renamed the Gemini Enterprise Agent Platform, while Google Agentspace, serving as an employee-focused AI assistant, is now part of the unified Gemini Enterprise product. Key announcements include a no-code agent builder for Google Workspace, a revamped developer platform featuring over 200 models, including third-party options like Anthropic’s Claude, a web-browsing agent named Project Mariner, managed MCP servers across Google Cloud services, and the production-ready Agent2Agent protocol for cross-platform agent communication. CEO Thomas Kurian labeled the keynote "The Agentic Cloud," explicitly comparing Google's integration to competitors who provide only components.

      The timing is intentional, considering that OpenAI's Operator is achieving an 87% success rate on complex browser task benchmarks while collaborating with Cognizant and CGI to promote its Codex coding agent among enterprise software vendors, with enterprise revenue now constituting 40% of OpenAI's total. Anthropic has established a marketplace for Claude-powered enterprise tools and its Model Context Protocol has been implemented on 10,000 servers with 97 million monthly SDK downloads. Google currently occupies third place in cloud market share, lagging behind AWS and Microsoft Azure, but finished the fourth quarter of 2025 with the fastest growth rate at 50% year over year. Google is betting that vertical integration, controlling the model, runtime, silicon, and distribution through Workspace, gives it a competitive edge that rivals cannot replicate.

      The agent stack

      Google Workspace Studio is the most prominent announcement aimed at consumers. It is a no-code platform that enables business users to create and deploy AI agents across Gmail, Docs, Sheets, Drive, Meet, and Chat using simple language. Users can input instructions like “every Friday, ping me to update my tracker,” allowing Gemini to generate the necessary automation. Workspace Studio integrates with third-party applications such as Asana, Jira, Mailchimp, and Salesforce, and can access external APIs through webhooks or execute custom logic using Apps Script. This feature is being rolled out for Google Workspace's business, enterprise, and education customers.

      The developer platform, now called the Gemini Enterprise Agent Platform, has received significant enhancements. The Agent Designer, a visual canvas for creating agent workflows, is currently in preview. Agent Engine Sessions and Memory Bank, which provide agents with persistent context across interactions, are now generally available. The new Agent Garden offers prebuilt agent solutions for customer service, data analysis, and creative tasks. A free tier via Express mode reduces access barriers. The Model Garden now hosts over 200 models, including Google's own Gemini and Gemma families, third-party models like Anthropic Claude, and open models such as Llama. Additionally, Google introduced six new agents for data engineering and coding in BigQuery, including a data engineering agent that automates pipeline creation from natural language inputs and a code interpreter that converts queries into executable Python with visualizations. Partner agents from companies like Box, Workday, Salesforce, ServiceNow, Dun and Bradstreet, and S&P Global are integrated into the platform, providing enterprises with prebuilt capabilities for various functions.

      Project Mariner, Google DeepMind’s web-browsing agent powered by Gemini 2.0, performs at an 83.5% score on the WebVoyager benchmark and manages ten concurrent tasks on cloud-based virtual machines. The agent automates shopping, information retrieval, and form completion, and is accessible to Google AI Ultra subscribers in the United States. Future developments include a visual builder named Mariner Studio in Q2, cross-device synchronization in Q3, and an agent marketplace in Q4.

      Perhaps the most strategically important announcement, though less visible to users, involves Google’s Agent2Agent (A2A) protocol, which has expanded from over 50 technology partners to 150 organizations in active use, managing real tasks among agents created on diverse platforms. This protocol is now overseen by the Linux Foundation’s Agentic AI Foundation and has reached version 1.2, implementing signed agent cards using cryptographic signatures for domain verification. Major firms such