Gemini Enterprise Agent Platform, formerly Vertex AI Agent Builder
Gemini Enterprise Agent Platform is Google Cloud's offering for designing, deploying and governing AI agents, and the successor to Vertex AI Agent Builder. It is aimed at technical teams in companies that already work on Google Cloud.
Data as of October 22, 2026
Key figures
| Developer | |
|---|---|
| Type | Low-code |
| Languages | Python, TypeScript, Go, Java, Kotlin |
| Open source | No |
| License | proprietary |
| First release | April 2024 |
| Suggested level | Advanced |
| Use cases | multi-agent, RAG, workflows, chatbot |
| Self-hostable | No |
| Business model | Paid |
| Cloud offering | Yes |
| Multi-agent | Yes |
| MCP support | Yes |
| Bring your own model | Yes |
| Official site | Official site · Documentation |
A new name for Google Cloud’s agent offering
Google introduced Vertex AI Agent Builder in April 2024. As of October 22, 2026, the official documentation refers to Gemini Enterprise Agent Platform, “an evolution of Vertex AI”, and the Agent Builder product page redirects to the new name. Several services were renamed: Vertex AI Agent Engine is now Agent Runtime, Vertex AI Studio is now Agent Studio, and Vertex AI Search is now Agent Search. Keep that mapping in mind when you follow older tutorials.
What you build, and how
The platform covers the full lifecycle of an AI agent, organized in four areas:
- Build: Agent Studio, a low-code visual canvas; the open-source Agent Development Kit for writing agents in code; Agent Garden for starting from samples; Model Garden for picking a model; RAG Engine and Vector Search for connecting your documents through RAG.
- Scale: Agent Runtime to host agents, with sessions, memory (Memory Bank) and Python code execution.
- Govern: Agent Registry, which catalogs agents, tools and MCP servers, plus Agent Identity, Agent Gateway and Model Armor for screening risky content.
- Optimize: agent evaluation and observability.
In practice, this is a tool for teams that want to industrialize agents, including systems where several agents work together, under the same security standards as the rest of their IT.
Cost and cloud lock-in
There is no flat subscription: everything is billed by usage. As of October 22, 2026, the pricing page lists $0.085 per vCPU-hour for running agents and $0.30 per GiB per month for storage; Sessions and Memory Bank have been billed since September 1, 2026. Model costs, measured in tokens, come on top. New Google Cloud customers get $300 in trial credit. Your budget is therefore hard to predict without a measured pilot.
Dependence on Google Cloud is real for runtime, governance and billing. Two things soften it: models from multiple providers, and the open-source ADK, whose agent code can run elsewhere.
Strengths, limits and alternatives
Strengths: an end-to-end chain from design to monitoring, and governance built for large organizations. Limits: naming that changes often, a bill with several components, and a steep learning curve for non-technical users.
On AWS, the counterpart is Amazon Bedrock AgentCore; in the Microsoft world, Copilot Studio. Official documentation: platform overview.
Frameworks of the same type
- Langflow · RAG, multi-agent, workflows
- Microsoft Copilot Studio · chatbot, workflows, automation
- Dify · chatbot, RAG, workflows
- n8n · automation, workflows, RAG
- Botpress · chatbot, RAG, automation
- Relevance AI · multi-agent, automation, workflows
Frequently asked questions
Does Vertex AI Agent Builder still exist?
The name has changed: Google now presents Gemini Enterprise Agent Platform as an evolution of Vertex AI, and the old Agent Builder product page redirects to it. Several components were renamed too, for example Vertex AI Agent Engine became Agent Runtime.
Do I need to code to use the platform?
Not necessarily to get started: Agent Studio is a low-code visual canvas. For more complex agents, Google relies on the Agent Development Kit, available in Python, TypeScript, Go, Java and Kotlin.
How much does Gemini Enterprise Agent Platform cost?
Billing is usage-based. As of October 22, 2026, running agents on Agent Runtime costs $0.085 per vCPU-hour and storage costs $0.30 per GiB per month, plus the tokens of the models you use. Sessions and Memory Bank have been billed since September 1, 2026.
Can I use models other than Gemini?
Yes. The platform's Model Garden offers models from Google and other providers, including Anthropic's Claude and the open Gemma models. Deployment, however, stays on Google Cloud.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.