GE Demo Generator
Instantly generate custom demo environments for any business domain. This tool dynamically synthesizes datasets and configures agents tailored to your client's unique challenges.
How GE Demo Generator Works
This web application does not deploy resources directly into your project. Instead, it dynamically synthesizes a personalized Setup Shell Script based on your custom inputs. Once you execute this script in your Google Cloud Shell, it automatically provisions all necessary resources (BigQuery, Firestore, Cloud Run) securely within your own Google Cloud environment, allowing you to instantly start demonstrations.
The synthesized agent uses a dual-model architecture: a Flash-Lite coordinator (root_agent) handles routine interactions at low latency, while a Pro sub-agent (deep_analysis_agent) is automatically delegated to for complex multi-step reasoning tasks. Models are configurable via --model-analysis-agent and --model-root-agent CLI flags.
Additionally, the accelerator generates domain-specific Sample Prompts, PDFs, and Excel files alongside the datasets. This allows you to enjoy a rich, living End-to-End pre-sales demonstration seamlessly.
Agent Runtime Architecture
The diagram below illustrates the dual-model delegation architecture. The Flash-Lite Root Agent coordinates most interactions directly, delegating complex multi-step analysis to the Pro Deep Analysis Agent. Both agents share the same MCP toolsets and image generation capabilities.
Deep Dive Resources
Need a full comprehensive explanation? Deep-dive into our architecture guidelines in the GitHub Repository.
Step 1: Define Business Challenge
Select a template or describe your customer's problem
Enter your customer's domain to automatically research their company, identify business challenges, and find agent-automatable workflows.
Business Challenges
Agent-Automatable Workflows
Click to select workflows for your scenario. Selected workflows will be included in the generated business challenge.
1. Select Industry
2. Filter by Job Function
3. Select a Template
Analyzing business goal...
Optimizing scenario with Gemini Flash Lite
Adds Google sign-in for the demo user WITHOUT the Developer-Preview Workspace MCP servers (no allowlist needed). With the Managed Agent enabled, deliverables can be saved to the user's Drive as native Google Slides / Docs / Sheets, and the autonomous agent acts on Gmail / Chat / Calendar by running the open-source Workspace CLI (gws) inside its sandbox — the CLI is pre-installed into the sandbox during setup warm-up, and its usage skills are mounted at /.agent/skills. Note: posting to Google Chat additionally needs a one-time Google Chat API app configuration in the demo project — the setup script tutorial (step 4) walks through it. Redundant when Workspace MCP is already enabled.
Deploy a fully autonomous Managed Agent (Antigravity harness) the demo agent can delegate long-running tasks to: web research, code execution, and professional deliverables (slides, documents, PDFs, web pages). Pre-GA Preview — works in any project with the Vertex AI Agent Platform API enabled (no allowlist required); agent provisioning adds ~10 min to setup. Combine with Workspace MCP or the Workspace Authorization toggle to unlock Google Drive / Workspace handoff. Enabled by default — switch it off for quick lightweight demos.
Let the agent operate a real Chromium browser (Gemini 3.5 Flash Computer Use) to run no-API websites. Watchable via a live-view link + in-chat screenshots.
logic synthesis & compilation
Domain Logic & Data Synthesis
Waiting...
Deployable Bundle Packaging
Waiting...
⚡ Dynamic Tip
Tips are loading...
Google Cloud Target Architecture
Live BlueprintRestoring demo assets...
Loading data, images, and agent config
Script Generated! What happens next?
Step 2: Review
Review the synthesized data and agent instructions below. You can refine them before proceeding to deployment.
Step 3: Deploy
Copy the setup script from Step 3 and paste it into Google Cloud Shell. The script will provision resources in your selected Google Cloud project, deploy the agent to Cloud Run, and register it in Gemini Enterprise.
Step 4: Run Live Demo
Once the script completes, open the Gemini Enterprise app to interact with your agent. To help you run an effective demonstration, use the generated demo prompts tailored specifically to your business challenge and the sample data.
Love this demo? Share it with your team!
Send a direct link so others can instantly load your demo
Step 2: Review
Step 3: Deploy
Run the synthesized environment in your own Google Cloud project
Environment Setup
💡 Recommended Environment
Click the button on the right to open Cloud Shell directly in your Google Cloud Console. We highly recommend using Cloud Shell as it comes pre-installed with all the necessary tools for this setup script. Alternatively, a step-by-step walkthrough is available by launching the Cloud Shell Tutorial.
1. Set Active Project ID
gcloud config set project [PROJECT_ID] && cd ~
2. Authenticate gcloud
gcloud auth login
gcloud auth application-default login
3. Agent Display Name (Optional)
This is how the agent appears in Gemini Enterprise. You can rename the human-readable part; the demo identifier in parentheses is kept fixed so --cleanup can remove this demo's resources reliably.
Execute & Confirm
Copy and paste the setup script into your Cloud Shell. And then hit the enter key.
Step 4: Run Live Demo
Engage with your agent using personalized scenarios
External Domain Files
Download these files and upload them to the agent chat if required
Confirm Generation
Ready to generate your demo environment?
Build Your AI Demo in Minutes
Create highly tailored, highly contextual Gemini agent demos instantly.
Core Value: Highly Tailored Custom Demos
Instantly generate highly tailored, highly contextual demos. The tool automatically synthesizes realistic sample data (e.g., handwritten faxes, order forms, and BigQuery master tables) and custom agent instructions specifically matching each customer's unique business and operational challenges.
Business Value: Win Pitches with A2UI
Show, don't tell. Demonstrate the ultimate value of Gemini Enterprise using A2UI (Agent-to-User Interface): the agent dynamically renders interactive buttons, custom data cards, and editable forms on the fly, providing a highly responsive, state-of-the-art Agentic AI UX for seamless human validation and approval.
Behind the Scene: Real Tech Stack
This is not a mockup. The generator spins up a fully-functioning cloud environment:
- Gemini & ADK: Advanced multi-step reasoning and tool orchestration.
- Rich MCP Ecosystem: Live connections to BigQuery, Google Maps, and Google Workspace.
- Interactive Data Viewer: A Flask Web UI to visually track documents and database changes.
How to Deploy: Automatic Cloud Provisioning
This tool generates a local setup script. Simply run it in your own Google Cloud Shell to automatically provision all demo resources (Cloud Run, BigQuery, Firestore) directly inside your Google Cloud project. All data remains securely within your own project boundary.
document.pdf
What's New
3 new capabilities to elevate your demos
Autonomous Long-Horizon Agents - Managed Agents API (Antigravity)
Your demo agent can now delegate long-horizon agentic tasks to a fully autonomous Managed Agent built on the Antigravity harness (Managed Agents API, Pre-GA). Working for minutes at a time in its own cloud sandbox, it researches the live web, queries the demo's BigQuery & Firestore data, writes and runs code, and crafts professional deliverables - presentation decks, documents, PDFs, and web reports - with live progress streamed into the chat and results announced automatically. Pair it with Google Workspace (via Workspace MCP or the new lightweight Workspace Authorization toggle) and the agent saves files into the user's Drive as native Slides / Docs / Sheets, drafts Gmail messages, posts to Chat, and books Calendar events with the user's own sign-in. Show truly realistic autonomous agentic work, end to end: research, build, deliver, and act across Workspace - all from one prompt. Now enabled by default for every new demo.
Knowledge Catalog for Metadata-Driven Insights
Every generated agent now taps Google Cloud Knowledge Catalog (Dataplex) through its managed MCP server. The agent semantically discovers the right data assets and reads rich business metadata - column meanings, units, allowed values, and table relationships - before it ever writes a query. Showcase governance-aware, metadata-grounded AI that picks the correct data and explains its reasoning, eliminating guessed columns and building instant trust in enterprise-grade analytics.
Custom MCP Server Import
Simply paste a GitHub URL to automatically analyze and integrate any Model Context Protocol (MCP) server into your demo. Instantly connect your demo agent to live, customer-specific data sources and tools, allowing you to build hyper-personalized scenarios that address exact customer pain points on the fly.