Guides / Prompting and shipping

How to Use Google AI Studio: Playground, Build Mode and Vibe Coding With Gemini

How to use Google AI Studio: test Gemini in the Playground, get an API key, vibe code an app in Build mode, and know when to move to a full-stack builder.

Mythex Team · 2026-11-01 · 6 min read

To use Google AI Studio, sign in at aistudio.google.com with a Google account, test prompts against Gemini models in the Playground, and click Get API key when you want to call Gemini from your own code. To vibe code an app, open Build mode, describe the app in a prompt, and Gemini generates it with a live preview that you refine by chatting. You can then deploy it to Cloud Run, sync it to GitHub or download the code.

This guide walks through each part as Google documents it at the time of writing (November 2026), then covers where AI Studio is strongest and when a builder with its own database and hosting is the better fit. Google updates AI Studio often, so check its docs for the latest names and limits.

What Google AI Studio is

AI Studio is Google's browser workspace for building with Gemini. It serves two kinds of people:

• Developers who want to try Gemini models, tune a prompt and copy working code into their own project.
• Builders who want to describe an app and have Gemini write it, using Build mode.

Google states that AI Studio usage is free of charge in the regions where it's offered, with limited free access to certain models. Paid API tiers, some models, heavy grounding with Google Search, and hosting on Google Cloud are billed separately.

Part 1: The Playground

The Playground opens by default with a new chat prompt. It's where you test what a model can do before building anything.

1. Pick a model and type a message in the input box, then click Run.
2. Open Run settings (the tune icon at the top right) to change model parameters and safety settings, and to switch on structured output, function calling, code execution or grounding.
3. Add System Instructions to set the model's role and tone, for example: "You are a friendly assistant for a bakery. Answer in two sentences and suggest one product."
4. Click Get code to turn the prompt and settings into code in your preferred language.

Besides chat, the Playground offers other prompt types such as realtime streaming and video generation.

Getting an API key

When you want Gemini inside your own app, create a key from the Get API key page (aistudio.google.com/apikey). Treat it like a password: keep it in your server's environment variables or secrets, never in code that runs in a visitor's browser. Our guide on keeping API keys safe explains why.

Part 2: Vibe coding in Build mode

Build mode is where "vibe coding with Gemini" happens. You describe an app, and AI Studio's agent writes the code and shows the result next to it.

Starting a project

You have several ways in:

• Type a prompt describing the app.
• Press I'm Feeling Lucky for a generated idea.
• Remix an app from the App Gallery with Copy App.
• Import from GitHub using the Add files menu.

You can add AI Chips to pull in features such as image generation or Maps data, and dictate prompts with speech-to-text.

What it builds

According to Google's documentation:

• Web apps by default: a React frontend plus a Node.js server, so API calls can stay on the server, npm packages install automatically, and the app can connect to databases.
• Native Android apps in Kotlin and Jetpack Compose, which you can preview in a browser emulator, install over ADB, and publish to Google Play for internal testing.
• Firebase setup done for you: Firestore for data and Firebase Authentication with Sign in with Google.
• Google Workspace integrations such as Gmail, Sheets, Docs, Drive and Calendar.

A first prompt

Build a recipe app. I paste a recipe or a link, and Gemini turns it into a clean card with ingredients, steps and cooking time. I can save recipes, search them and scale ingredients for a different number of servings. Let me sign in with Google so my recipes are saved to my account.

This plays to AI Studio's strengths: Gemini is part of the feature, sign-in is Google, and storage is Firebase.

Refining the app

• Chat with the agent: "Add a shopping list that combines ingredients from the recipes I tick."
• Annotation mode: highlight part of the app's screen and describe the change you want there. It's faster than describing which button you mean.
• Code tab: view the generated code and edit it directly if you want to.

Google's docs say Build mode keeps the Gemini API key as a server-side secret (GEMINI_API_KEY, visible in the Secrets panel) rather than putting it in client code. Android apps use Firebase AI Logic and App Check instead of a key.

Sharing and deploying

• Sharing: apps are private by default. People you share with can view and fork the code, and those with edit rights can change it. Gemini calls made by people using your shared app count toward your usage, and AI Studio warns you before sharing if costs are likely.
• Cloud Run: deploy from AI Studio as a scalable service. Google Cloud pricing may apply.
• GitHub: two-way sync with a repository, with a conflict view when both sides changed.
• ZIP download: take the code anywhere. You'll set the API key on your own host.

Tips for better results in Build mode

• Lead with what Gemini does in the app. Build mode is at its best when the AI is the feature: summarising, extracting, generating images, answering questions over a document.
• Say what's saved and for whom. "Each user's recipes are private to them" is a rule worth stating up front.
• Change one thing per message, and test it in the preview before the next.
• Review the Firebase security rules before real users arrive. The app working for you doesn't mean other users can't read your data. Ask: "Show me the Firestore rules and explain who can read and write each collection."

For more prompts that work in any tool, see our vibe coding prompts cheat sheet.

Where Google AI Studio is strongest

• Apps where Gemini is the core feature, with the key kept on the server.
• Google-centred stacks: Firebase, Sign in with Google, Workspace data, Maps.
• Developers who want a quick prototype that they then sync to GitHub and keep building.
• Native Android experiments, previewed in the browser.

When to move to a builder with database and hosting

AI Studio is a strong tool, and for many projects it's enough. It's worth considering a full-stack AI app builder when:

• You want a relational database. Firestore is a document database. If your app has customers, orders and invoices with relationships between them, a Postgres database can be simpler to reason about. Our Supabase vs Firebase guide explains the difference.
• You don't want to manage a Google Cloud project. Deploying to Cloud Run means a Cloud account, billing and its settings. Some people would rather publish with one click and have hosting included.
• You want hosting, database and storage on one bill. On Mythex, these are paid from monthly Cloud credits included in the plan.
• You need iPhone as well as Android. On Mythex Pro, a published app can become an Android app you upload to Google Play and an iPhone app sent to TestFlight with your own App Store Connect key, without a Mac. (These open your published app; they aren't native Swift or Kotlin code.)
• The AI isn't the point of the app. A booking system or client portal doesn't need Gemini at its centre, and a general builder may fit more naturally.
• You want your own domain without configuring Cloud services.

Mythex works like this: describe the app in chat, in any language, and it builds a real full-stack app with a live preview, attaches a Postgres database and file storage, and publishes to a public URL in one click. The Free plan needs no card and includes publishing; Pro is $25 a month and adds code editing, export, GitHub, custom domains and phone apps. See the AI app builder page.

Either way, the code is yours to take with you: AI Studio through GitHub or ZIP, Mythex through export and GitHub on Pro. For a wider view of the options, see the best vibe coding tools and ChatGPT vs Claude vs Gemini for building apps.

Google AI Studio の使い方（要点）

1. aistudio.google.com に Google アカウントでログインします。
2. Playground でモデルを選び、プロンプトを入力して Run を押します。Run settings でパラメータや機能を切り替えられます。
3. 自分のアプリから Gemini を使うときは Get API key でキーを作成し、サーバー側の環境変数に保存します。
4. アプリを作るなら Build mode を開き、作りたいアプリを文章で説明します。プレビューを見ながらチャットや注釈モードで修正します。
5. 完成したら Cloud Run にデプロイするか、GitHub と同期するか、ZIP でダウンロードします。

データベース、ホスティング、独自ドメイン、iPhone と Android のアプリを一か所で管理したい場合は、Mythex のようなフルスタックの AI アプリビルダーも選択肢になります。