13 Hands-On Demos for Building with the Gemini Enterprise Agent Platform

Google Cloud AI has released 13 hands-on demos showing how to build, scale, govern, and optimize AI agents with the Gemini Enterprise Agent Platform.

Each demo teaches concepts, patterns, and architectures that can be used right away. Installing the Agents CLI on a coding agent (Claude Code, Codex, etc.) automatically adds 7 skills, so describing the desired functionality in natural language lets the agent handle scaffolding, evaluation, deployment, and monitoring automatically.

Build (4 demos): Start with the ADK-based fundamentals. Building your first agent, approval workflows and human-in-the-loop, connecting data with the Model Context Protocol (MCP), and dynamic UI rendering with Agent-to-UI (A2UI).

Scale (4 demos): Getting ready for production environments. Deploying with Agent Runtime, long-running agents that persist for over a week, stateful data science agents, and building an end-to-end dashboard.

Govern (2 demos): Managing agents across the organization. Strengthening security with TDD and the STRIDE threat model, and runtime access control with Agent Gateway plus injection defense with Model Armor.

Optimize (3 demos): Improving agent quality. Evaluating prompt improvements with AutoRaters, multi-language agent pipelines based on Python/Go, and orchestration integrating various frameworks such as CrewAI, LangGraph, and ADK.

All demos leverage the graph-based workflow API of Agent Development Kit (ADK) 2.0, with Cloud Trace, Cloud Logging, and BigQuery analytics automatically integrated.