Google launches Gemini Enterprise Agent Platform with Agent Studio and 200+ models

• Google positioned Gemini Enterprise Agent Platform as the next layer on top of Vertex AI, and GoogleDeepMind's launch post says it is meant to cover the full lifecycle: build, scale, govern, and optimize enterprise agents.
• According to GoogleDeepMind's platform diagram, the stack combines new surfaces like Agent Development Kit, Agent Studio, Agent Gateway, Agent Registry, Agent Evaluation, and Agent Optimizer with GA runtime pieces such as Agent Runtime, Agent Sessions, Agent Sandbox, and Agent Memory Bank.
• GoogleDeepMind's thread says the platform exposes 200-plus models through Model Garden, including Gemini 3.1 Pro, Gemini 3.1 Flash Image, Lyria 3, and open models like Gemma 4.
• On the app side, Google's Gemini Enterprise feature list adds Inbox, Canvas editing for Docs and Slides, reusable skills, projects, long-running agents, and an Agent Designer for employees building scheduled or trigger-based agents.
• A separate rollout around AI Studio lets Google AI Pro and Ultra subscribers use AI Studio with higher limits, but OfficialLoganK's reply said Workspace accounts are not supported yet after cedric_chee's screenshot surfaced the gap.

You can jump straight to Google's announcement page, inspect GoogleDeepMind's architecture diagram, and compare it with the Gemini Enterprise app infographic, which quietly turns the story from a Vertex AI platform launch into a broader workplace agent rollout. There is also a separate Workspace Intelligence post, plus testingcatalog's early roundup calling out the open partner ecosystem and Agent Gallery before Google's main thread spread.

The cleanest read is that Google is renaming and widening the old Vertex AI agent story, not starting from zero. GoogleDeepMind's wording calls the platform "the evolution of Vertex AI," while Google's main account frames the problem as managing thousands of agents rather than merely building one.

That platform pitch bundles four jobs into one control plane:

• Build: Agent Development Kit, third-party agent frameworks, Agent Studio, Agent Garden, Gemini API, and Model Garden, per GoogleDeepMind's architecture diagram.
• Scale: Agent Runtime, Agent Sessions, Agent Sandbox, and Agent Memory Bank, all marked GA in the diagram screenshot.
• Govern: Agent Gateway, Agent Identity, Agent Registry, Agent Anomaly Detection, Model Armor, Agent Policy, Agent Security, and Agent Compliance, according to GoogleDeepMind's platform map.
• Optimize: Agent Evaluation, Agent Simulation, Agent Observability, and Agent Optimizer, again from the platform map.

The notable bit is how much of the diagram is agent-specific infrastructure rather than model branding. Google's own materials keep coming back to identity, registry, gateway, policy, anomaly detection, evaluation, and observability, which is a much more operations-heavy story than a simple model release.

Google says the platform gives enterprises access to more than 200 models through Model Garden, and GoogleDeepMind's thread explicitly names Gemini 3.1 Pro, Gemini 3.1 Flash Image, Lyria 3, and Gemma 4.

The diagram in GoogleDeepMind's post makes that model layer look deliberately mixed:

• Gemini models
• Third-party and open models
• Model training
• Model inference
• Tools, data, and other agents connected through A2A, grounding, RAG, MCP, search, APIs, connectors, A2UI, AP2 and UCP, and Cloud Marketplace

That matters because Google's agent pitch is not locked to Gemini-only execution. testingcatalog's early summary also highlights an open partner ecosystem with third-party agents from Oracle, Salesforce, and ServiceNow, which matches the platform diagram's emphasis on connectors and external frameworks.

The developer-facing launch sits next to a second layer: a Gemini Enterprise app meant for discovery, sharing, and day-to-day use of agents inside the workplace.

From Google's feature infographic, the app-side additions are:

• Inbox: a central place to monitor, guide, and manage agent activity.
• Canvas: Docs and Slides editing in one pane, with Microsoft 365 interoperability called out in the screenshot.
• Long-running agents: autonomous multi-step workflows that can run for hours to days in Google's cloud sandboxes.
• Projects: a shared workspace where humans and agents work with common context across Workspace, OneDrive, chats, and more.
• Agent Designer: a no-code or low-code builder for schedule-based and trigger-based agents using enterprise connectors.
• Security and governance foundations: Agent Identity, Agent Registry, and Agent Gateway surfaced inside the app experience.

That creates a split architecture. The platform side looks like infrastructure for developers and IT, while the app side looks like the distribution layer where employees discover, run, and supervise those agents.

Google is also attaching a broader context layer to this rollout. Google's Workspace Intelligence post says Workspace Intelligence is meant to understand a company's work ecosystem across Workspace apps, with business context, work-style personalization, and privacy controls built in.

The tweet only gives a high-level description, but it is enough to show where Google wants the enterprise agent story to land: not just agents as isolated tools, but agents grounded in workspace data, documents, datasets, and app context. The related post is linked as Workspace Intelligence.

A separate product change landed just before Cloud Next: AI Studio now works with Google AI Pro and Ultra subscriptions. testingcatalog's screenshot shows Google pitching those plans directly inside AI Studio, with higher limits and access to more models and agents.

That rollout still has account-boundary caveats. After cedric_chee's screenshot showed the upgrade flow failing for a Google Workspace AI Expanded Access account, OfficialLoganK's reply said there was "No workspace support yet." demishassabis's repost and algo_diver's repost of Google's announcement show Google amplifying the consumer-subscription path anyway.

The result is slightly messy but revealing. Google now has an enterprise agent platform, a Gemini Enterprise app, Workspace Intelligence, and a subscription-backed AI Studio path all moving at once. The pieces line up around agents, but they are not yet one entitlement story.