Google Gemini Platform (Agent Builder)

Looking at Google Cloud for building AI agents? Here's a heads-up first. If you go to Google Cloud's Agent Builder product page today, it's called the Gemini platform. Here's what that page says (we checked it on Oct 28, 2026).

Google describes it as an open platform for building, scaling, governing, and improving AI agents grounded in your own business data. It's aimed mostly at technical teams.

Here's what's in it, in Google's words:

• Agent Studio: design, test, and manage prompts for Gemini models using plain language, code, images, or video.
• Model Garden: 200+ Google and third-party models and tools, including Google's Gemini models, Anthropic's Claude family, and open models like Gemma.
• Agent Development Kit (ADK): a developer toolkit for building and fine-tuning more advanced agents.
• Google Antigravity: a desktop app and CLI to steer and orchestrate several agents at once. Google's example is a product launch, with agents handling website code, on-brand assets, and customer emails.
• Agent Search: document retrieval for your agents.

There are also tools for data scientists, like notebooks, model training, and pipelines.

Pricing, as listed: it's pay-as-you-go, based on what you use. A few specifics from the page:

• New Google Cloud customers get $300 in free credits, valid for 90 days.
• There's a monthly free tier for agent runtimes: the first 50 hours of compute, 100 hours of memory, and 1 GiB of storage.
• Text, chat, and code generation starts at $0.0001 per 1,000 characters.
• Agent Search document retrieval is about $4.00 per 1,000 queries, plus storage.
• Pipelines start at $0.03 per run.

Model prices vary by model and tier, and Google suggests contacting sales for detailed estimates. So exact costs for your use are unknown until you price it out.

A quick note: the customer quote and report on Google's page are its own marketing, so we've left them out.