13 Hands-On Demos for Gemini Enterprise Agent Platform

Earlier this year, Google unleashed the Gemini Enterprise Agent Platform—a swamp-ready toolkit for building, scaling, governing, and optimizing AI agents. Now they’ve dropped 13 hands-on demos that take you from zero to production faster than a capybara sliding into a cool pool. The best part? You don’t have to follow them step-by-step. Just install Agents CLI into your favorite coding agent (Antigravity, Claude Code, Codex, whatever floats your boat) and describe what you want in plain English. Your coding agent becomes an instant ADK expert, scaffolding, evaluating, deploying, and monitoring your agent without ever leaving your editor. Let’s dive into the murky waters!

Build AI Agents

These demos are built on the code-first ADK, starting at the foundation and working up. Perfect for capybaras who like to build from the ground up.

• 1. Start here: build your first agent with ADK. The ADK Foundation codelab is your perfect on-ramp. Set up your environment, define a basic conversational agent powered by Gemini, configure settings, and test it via CLI and web UI. If you’ve never touched ADK, start here.
• 2. Build an event-driven approval agent with human-in-the-loop. The ambient expense agent codelab is the most complete demo. You build a corporate expense agent using ADK 2.0’s graph-based workflow. Expenses under a threshold auto-approve; anything above goes through PII redaction, prompt-injection defense, Gemini compliance analysis, and pauses for human review. Mounted behind FastAPI, triggered from Pub/Sub, and graded with LLM-as-judge.
• 3. Connect agents to your data with MCP. The MCP codelab shows how to build reusable MCP tools that let Gemini query BigQuery, search files, and call APIs. MCP is open, so your tools work across vendors.
• 4. Build a dynamic frontend with Agent-to-UI (A2UI). The A2UI codelab teaches you to build an agent that renders real UI components (layouts, charts, menus) that update dynamically as the conversation flows. The agent assembles the UI on the fly.

Scale AI Agents

A prototype on your laptop is like a capybara in a kiddie pool—cute, but not ready for the open swamp. These demos handle production traffic, memory, and orchestration.

• 5. Deploy a stateful data science agent to Agent Runtime. The Stateful Data Science Agent codelab walks you through building a BigQuery agent that remembers user preferences via Memory Bank, then deploys to Agent Runtime. Infrastructure, scaling, and session management are handled automatically.
• 6. Build long-running agents that pause, resume, and never lose context. This tutorial covers durable state machines, event-driven idle handling, and checkpoint-and-resume with persistent sessions. Example: an onboarding coordinator that survives container restarts.
• 7. Deploy the expense agent to Agent Runtime with Agents CLI. The Deploy to Agent Runtime codelab takes the expense agent to production. Scaffold config with CLI, dry run, then deploy. Cloud Trace, Logging, and BigQuery Agent Analytics wire in automatically.
• 8. Give your production agent a real front end. The frontend codelab builds a manager dashboard on Cloud Run, connected to Agent Runtime via OIDC-authenticated Pub/Sub. Managers can resume human-in-the-loop sessions from the browser.

Govern AI Agents

Scaling across an organization requires guardrails. These demos keep your agents from wandering into dangerous waters.

• 9. Secure your agent’s lifecycle from the first commit. The Secure Agentic Coding codelab shows test-driven development, STRIDE threat modeling, Semgrep pre-commit hooks, and PreToolUse gates that block risky actions. Plant a hardcoded API key and watch the agent catch it.
• 10. Control agent access with Agent Gateway. The Agent Gateway codelab covers runtime governance: deploy a multi-tool ADK agent on Agent Runtime that calls MCP servers through Agent Gateway with end-to-end mTLS, IAP auth, IAM authorization, and Model Armor for content inspection.

Optimize AI Agents

Shipping is just the start. These demos help you close the loop on quality.

• 11. Drive the agent quality flywheel from your coding agent. This tutorial introduces a five-stage evaluation flywheel: prepare data, run inference, grade with Google’s adaptive AutoRaters, analyze failures, and optimize. AutoRaters are built on principles Google uses for its own models.
• 12. Build a cross-language multi-agent pipeline with A2A. This tutorial walks through a contract compliance pipeline: a Python agent extracts terms, a Go agent validates them, connected via A2A and orchestrated by ADK.
• 13. Scale agents across frameworks with CrewAI, LangGraph, A2A, and ADK. The codelab shows an ADK control room delegating to LangGraph state machine, which dispatches to CrewAI execution crew, all via A2A. If one step fails, the control room re-plans automatically.

Get Started

If you want to see the full agent development lifecycle in under 10 minutes, watch this walkthrough. Otherwise, install Agents CLI, open your coding agent, and start building today. For more on cloud platforms, check out our Google Cloud review and Vercel review.

Original announcement published on Google Cloud.