Google has launched Gemini Enterprise Agent Platform, a new Google Cloud service that repositions Vertex AI as a broader system for building, governing, and running business agents.

What Google announced

In its Cloud Next posts, Google described the platform as a single place for technical teams to build, scale, govern, and optimize agents. The launch bundles model access, agent tooling, runtime, security, and observability into one product. Google also said future Vertex AI services and roadmap updates will flow through Agent Platform instead of a separate standalone Vertex AI track.

Google's own materials say the platform includes access to first-party models such as Gemini 3.1 Pro, Gemini 3.1 Flash Image, and Lyria 3, while also supporting third-party models from Anthropic, including Claude Opus, Sonnet, and Haiku. In a separate Cloud Next keynote post, Google said support is also being added for Claude Opus 4.7.

Why it matters

The notable signal is not just another agent builder. Google is explicitly splitting the stack between technical teams and business users. Agent Platform is aimed at developers and IT teams managing identity, registry, gateway, runtime, and evaluation, while the Gemini Enterprise app exposes no-code tools such as Agent Designer, Inbox, and long-running agents to employees.

That makes this a platform move, not just a demo feature drop. The conservative takeaway is that Google wants enterprises to treat agent operations more like governed software infrastructure, while still letting customers mix Gemini with outside frontier models when they deploy it.