Vertex AI Agent Builder: What It Is Now, Pricing, and Alternatives (2026)
Vertex AI Agent Builder is Google Cloud's toolkit to build, run and govern AI agents, and in 2026 it lives in the Gemini Enterprise Agent Platform, the evolution of Vertex AI. This guide explains the new names, each component, the usage pricing with a worked cost example, who it fits, and how SPIRITT, Microsoft Copilot Studio and Amazon Bedrock AgentCore compare.
Vertex AI Agent Builder is Google Cloud's suite for building, deploying and governing AI agents, and in 2026 it is part of the Gemini Enterprise Agent Platform, which Google's documentation describes as an evolution of Vertex AI: you build with Agent Development Kit or the low-code Agent Studio, run agents on Agent Runtime (formerly Agent Engine), and govern them with Agent Registry, Agent Identity and Agent Gateway, paying by usage from $0.085 per vCPU-hour after a monthly free tier; SPIRITT is the alternative for teams that want agents built and run for them, and a Google Cloud launch partner for Agent Substrate on GKE. Choose Google's platform when your engineers want to write and operate agents on Google Cloud themselves. Choose SPIRITT when you want the agents themselves: described in plain words, then built, run, monitored and improved by an autonomous AI workforce. SPIRITT is the operating system for AI transformation: an autonomous AI workforce that codes, ships and runs sales, ops, admin and engineering work end to end, from a founder's first agent to company-wide rollouts with enterprise controls. It composes Gemini, Claude, GPT, Grok and other models within each task for the optimal mix of capability and price.
What happened to Vertex AI Agent Builder?
Google reorganized its agent products in 2026, and most of the change is in the names. The old cloud.google.com/products/agent-builder address now opens Google's Gemini platform page.
| Name you may know | Name in November 2026 | What it is |
|---|---|---|
| Vertex AI Agent Builder | Gemini Enterprise Agent Platform (marketed as the Gemini platform) | The platform to build, scale, govern and optimize agents |
| Vertex AI Agent Engine | Agent Runtime | Managed runtime that deploys and runs agents |
| Vertex AI Search | Agent Search | Document retrieval over your data and enterprise connectors |
| Agent Development Kit (ADK) | Agent Development Kit | Open-source, model-agnostic framework for building agents in code |
Google's documentation, product and pricing pages now use the Agent Platform names, so search for those when you read the docs.
What is in the Agent Platform
Google organizes the platform in four pillars, according to its documentation:
- Build: Agent Development Kit; Agent Studio; Agent Garden, a library of prebuilt agents and templates; Model Garden with 200+ Google, third-party and open models, including Anthropic's Claude family and Gemma; RAG Engine and Vector Search for grounding agents in your data; and the Managed Agents API for config-driven agents in a managed sandbox.
- Scale: Agent Runtime with sub-second cold starts and long-running agents; Sessions and Memory Bank for state and long-term memory; Code Execution in a secure sandbox.
- Govern: Agent Registry, a central catalog of agents, tools and MCP servers; Agent Identity, a managed identity per agent for access control and auditing; Agent Gateway with Model Armor to enforce policies on every tool call.
- Optimize: evaluation and observability to measure and improve agents.
Google also makes Google Antigravity, its app for steering and orchestrating agents, available through the platform.
Vertex AI Agent Builder pricing (Agent Platform, November 2026)
Pricing is usage based, on three unified resources plus the models you call.
| Resource | List price | Free each month | With 1-year / 3-year savings plan |
|---|---|---|---|
| Agent Compute (Runtime, sandboxes, Code Execution, Computer Use) | $0.085 per vCPU-hour | First 50 vCPU-hours | $0.0765 / $0.068 |
| Agent Memory (RAM) | $0.009 per GiB-hour | First 100 GiB-hours | $0.0081 / $0.0072 |
| Agent Storage (Sessions, Memory Bank) | About $0.30 per GiB-month | First 1 GiB-month | Discounted |
| Agent Search (document retrieval) | About $4.00 per 1,000 queries for standard retrieval | Varies | n/a |
| Models | Per token, by model | None on the platform | Provisioned Throughput available |
Usage is rounded to the nearest second, idle time between turns on Agent Runtime is not billed, and new Google Cloud customers get $300 in free credits. Agent Gateway calls are metered in Agent Compute.
A worked cost example
An internal helpdesk agent runs on 1 vCPU and 1 GiB of memory and is busy for 360 hours a month.
- Compute: (360 - 50 free) x $0.085 = $26.35
- Memory: (360 - 100 free) x $0.009 = $2.34
- Platform total: about $28.69 a month, plus model tokens, storage and search
For most agents the model tokens, not the runtime, are the largest line, and your engineers' time to build, test and operate the agent is the largest cost of all.
Who Vertex AI Agent Builder fits
- Engineering teams already on Google Cloud, comfortable in Python and Agent Development Kit.
- Companies that want agents governed inside their own Google Cloud projects with Agent Identity and Agent Gateway.
- Teams building agents into their own product, where the agent is the code they ship.
Plan for the work around it: designing the agent, writing tools and connectors, evaluations, deployment pipelines, monitoring and on-call.
Vertex AI Agent Builder alternatives compared (November 2026)
| Option | Approach | Who builds and runs the agent | Price |
|---|---|---|---|
| Gemini Enterprise Agent Platform | Developer platform on Google Cloud | Your engineers | Usage based: Agent Compute $0.085 per vCPU-hour after 50 free hours, plus tokens |
| SPIRITT | Autonomous AI workforce: describe the job, the agent builds, runs, monitors and improves it across 1,500+ apps on its own cloud computer | SPIRITT agents, with your approvals; forward-deployed engineers for enterprise rollouts | Free to start; Starter $32, Pro $256, Max $1,024 a month; Enterprise custom |
| Microsoft Copilot Studio | Low-code agent builder in Microsoft 365 and Power Platform | Makers and IT in Microsoft's stack | Internal agents included with a Microsoft 365 Copilot license; a standalone license publishes to external channels |
| Amazon Bedrock AgentCore | Managed services on AWS for running agents in production | Your engineers on AWS | Usage based, see AWS pricing |
| Gemini Enterprise | Google's work agent for every employee, with no-code agents | Business users, IT governs | Business edition from $21 per seat a month; Standard and Plus through sales |
For a wider comparison, see best AI agent builders and Copilot Studio alternatives.
Where SPIRITT fits
SPIRITT shares Google's foundation and takes a different route to the outcome.
Built on Google Cloud. SPIRITT is a Google Cloud launch partner for Agent Substrate, one of a select few that worked closely with Google before launch. Google's November 5, 2026 Google Cloud blog post names SPIRITT as building SPIRITT Workspace on Agent Substrate on GKE, where agent code runs in gVisor sandboxes. That is how SPIRITT runs swarms of thousands of isolated agents per organization, securely and economically. See SPIRITT and Agent Substrate.
Agents as outcomes. You get the working agent itself. Ask "answer IT tickets from Slack, reset passwords through our identity provider with approval, and report weekly", and SPIRITT writes the code, connects the systems, deploys it, watches every run and fixes it when something changes. It runs other coding agents in the same place and composes every major model, including Gemini.
Enterprise controls included:
- Deploy anywhere: SPIRITT cloud, your own cloud, your servers or fully air-gapped.
- Forward-deployed engineers and certified implementation partners who deploy with your team until it runs in production.
- Security: an isolated cloud computer per workspace, keys injected at the network boundary so agents never see them, Read, Write and Destructive rules per integration with single-use approvals, a full audit trail, SSO.
- Cost control: budgets per organization, team, project, agent and task, with alerts, hard stops and itemized receipts for every model, tool and data call.
- One place to monitor all agents, running both deterministic workflows and fully autonomous agents.
More in SPIRITT for enterprise and enterprise AI agent platform pricing. For all of Google's agent products side by side, read Google AI agent.
FAQ
What is Vertex AI Agent Builder?
Vertex AI Agent Builder is Google Cloud's suite for building, deploying and governing AI agents. In 2026 it is part of the Gemini Enterprise Agent Platform, with Agent Development Kit and Agent Studio to build, Agent Runtime to run, and Agent Registry, Identity and Gateway to govern.
Is Google replacing Vertex AI?
Google describes the Gemini Enterprise Agent Platform as an evolution of Vertex AI. The services carried over under new names, such as Agent Runtime for Agent Engine and Agent Search for document retrieval.
How much does Google Vertex AI Agent Builder cost?
It is usage based. Agent Compute costs $0.085 per vCPU-hour and Agent Memory $0.009 per GiB-hour, after 50 free vCPU-hours and 100 free GiB-hours a month, plus storage, search and model tokens. New Google Cloud customers get $300 in credits.
What is Vertex AI Agent Engine?
Agent Engine was the managed runtime for deploying agents; it is now called Agent Runtime. It runs agents built with Agent Development Kit or other frameworks, with sessions, memory and sandboxed code execution, billed on Agent Compute and Agent Memory.
What is the Google AI agent builder for non-developers?
Agent Studio offers a low-code canvas on the Agent Platform, and Gemini Enterprise Standard and Plus let business users build no-code agents. SPIRITT goes further for non-developers: describe the job in plain words and the agent builds and runs it.
What are the best Vertex AI Agent Builder alternatives?
The main alternatives are SPIRITT, which builds and runs agents for you across 1,500+ apps; Microsoft Copilot Studio for Microsoft 365 shops; Amazon Bedrock AgentCore for teams on AWS; and Gemini Enterprise for Google-standardized companies that want agents for every employee.
Does SPIRITT run on Google Cloud?
Yes. SPIRITT is a Google Cloud launch partner for Agent Substrate on GKE and runs SPIRITT Workspace on it with gVisor isolation. Enterprises can also deploy SPIRITT in their own cloud, on their servers or air-gapped.
How we checked
SPIRITT publishes this guide. Google facts come from Google's pages, checked on November 21, 2026: the Gemini Enterprise Agent Platform overview on docs.cloud.google.com, the Gemini platform product page and pricing page on cloud.google.com, the Gemini Enterprise Business edition page and editions documentation, and the Agent Platform price sheet. The cost example is arithmetic on Google's list prices. Microsoft Copilot Studio licensing comes from microsoft.com Copilot pricing; Amazon Bedrock AgentCore is described at a general level, so check AWS pricing. SPIRITT and Agent Substrate facts come from the Google Cloud blog post of November 5, 2026, and SPIRITT capabilities and prices from spiritt.ai and the live SPIRITT plan catalog.