Google AI Studio for Coding (2026): Build Mode, Gemini 4, Pricing, and When to Use an Autonomous Agent
How to use Google AI Studio for coding in November 2026: Build mode and its Antigravity Agent, full-stack web and native Android apps, GitHub sync, publishing to Cloud Run, the free tier and Gemini API prices. Plus how it compares with Antigravity, Gemini CLI, Lovable and Replit, and where SPIRITT, an AI agent that codes, ships and runs the business work around the app, fits.
Google AI Studio for coding means using Google's free, browser-based studio to write and test code with Gemini models, and above all its Build mode, where you describe an app in plain words and the Antigravity Agent generates a full-stack web app (React front end, Node.js server) or a native Android app (Kotlin and Jetpack Compose), with a live preview, GitHub sync and one-click publishing to Cloud Run. It is the fastest free way to prototype with Gemini. When the app has to connect to your business systems, run on a schedule and keep working after launch, SPIRITT is an AI agent that codes, ships and runs it end to end on its own cloud computer, composing Gemini with Claude, GPT and other models inside each task. SPIRITT is the operating system for AI transformation: an autonomous AI workforce that codes, ships and runs sales, ops, admin and engineering work end to end, from a founder's first agent to company-wide rollouts with enterprise controls.
Short version: use Google AI Studio to prototype and to get a Gemini API key, use Antigravity or Gemini CLI if you code in your own editor, and use SPIRITT when you want the finished, deployed system plus the operations around it.
What Google AI Studio does for coding (November 2026)
Google AI Studio (aistudio.google.com) is Google's web workspace for building with the Gemini API. For coding it has three parts:
- Build mode ("vibe coding"). Describe the app, pick web or Android, and AI Studio generates the files with a live preview on the right. You can start from a prompt, remix an app from the App Gallery, or import a project from GitHub.
- Chat and prompt testing. Try Gemini models on code questions, compare outputs and tune system instructions before you call the API.
- API keys and code export. Get a Gemini API key and copy working code for your own project.
Build mode in detail
According to Google's AI Studio documentation, Build mode now runs on the core of the Antigravity Agent, the agent harness from Google Antigravity. It keeps the context of your whole project, manages changes across many files and verifies code updates. What it builds:
- Full-stack web apps: a React front end by default and a Node.js server runtime, with npm packages installed as needed and secrets kept server-side.
- Native Android apps: Kotlin and Jetpack Compose, previewed in a browser emulator, installable on a device and publishable to Google Play for internal testing.
- Built-in services: Firebase Firestore and Firebase Authentication (Sign in with Google) set up automatically, plus Google Workspace APIs such as Gmail, Sheets, Drive and Calendar with OAuth handled for you.
- Real-time multiplayer apps through the server runtime.
You keep iterating in the chat panel, edit code directly in the Code tab, or use annotation mode to highlight part of the UI and describe the change.
From prototype to production
- GitHub: two-way sync, with AI-written commit messages, or a ZIP download to continue in your editor.
- Publishing: apps publish to Cloud Run with a shareable URL. Eligible personal accounts can publish up to two apps on the Google Cloud Starter Tier without a billing account; beyond that you link a paid Google Cloud project. Custom domains can be bought and connected from the Publish panel.
How to use Google AI Studio for coding: a 7-step workflow
- Write the brief first. Users, the three main screens, the data you store and who can see what. Paste it as your first prompt.
- Start in Build mode and choose web or Android.
- Ask for the data model before the UI. "Create the Firestore collections and security rules first." Apps go wrong at the data layer, not the buttons.
- Add sign-in early with Firebase Authentication, so every later feature respects users.
- Use annotation mode for UI tweaks instead of long text descriptions.
- Connect GitHub once the first version works, so every change is versioned and reviewable.
- Publish to Cloud Run, test the live URL on your phone, then add a custom domain.
Tip: keep secrets in Settings, never in the prompt, and review generated security rules before sharing a link.
Google AI Studio pricing for coding
The studio itself is free with a Google account. Costs come from the models your app calls and from hosting:
| Model (Gemini API) | Free tier | Paid tier per 1M tokens (input / output) |
|---|---|---|
| Gemini 4 Carbon (most capable, for complex software engineering) | Not available | $2 / $10 through November 30, 2026; $4 / $20 from December 1, 2026 |
| Gemini 4 Hydrogen (everyday coding and agents) | Free of charge | $1.50 / $7.50 |
| Gemini 3.8 Flash (long-horizon coding and agents) | Free of charge | $0.75 / $3.75 through December 31, 2026 |
The free tier has rate limits and Google may use free-tier content to improve its products; the paid tier does not. Batch API calls cost 50% less. Published apps on Cloud Run follow Google Cloud pricing outside the Starter Tier.
Google AI Studio vs other AI coding tools (November 2026)
| Tool | Best for | Where the code runs | Price (US) |
|---|---|---|---|
| Google AI Studio (Build mode) | Fast prototypes with Gemini; web and Android apps | Browser; publishes to Cloud Run | Free studio; Gemini API pay as you go |
| Google Antigravity | Agent-first coding in Google's desktop app, CLI and IDE extensions | Your machine, with background agents | Free; more quota with Google AI Pro $19.99 or Ultra $99.99 or $199.99 a month |
| Gemini CLI | Terminal coding agent | Your machine | Free and open source |
| Lovable | Polished web apps from prompts | Lovable cloud | Free, Pro from $25 a month |
| Replit | Apps built and hosted in one place | Replit cloud | Core $20 a month |
| SPIRITT | Apps plus the business work around them, built, deployed and operated | Its own isolated cloud computer per workspace | Free to start; Starter $32, Pro $256, Max $1,024 a month |
Google's async coding agent Jules winds down on December 31, 2026, so new projects should start in AI Studio, Antigravity or Gemini CLI.
Where an autonomous agent fits
Google AI Studio is excellent at the first 80% of an app: the screens, the data model, a working preview. The last 20% is where most projects stall, and it is mostly not code generation:
- Connecting real systems. Your CRM, accounting tool, helpdesk or a legacy portal with no API.
- Running on a schedule. Nightly syncs, weekly reports, reminders that fire without you.
- Watching it after launch. Errors, slow pages, a broken integration at 2 am.
- The business work around the app. Onboarding emails, a pricing page, support replies, the spreadsheet finance asks for.
SPIRITT handles that part. It is a frontier coding agent that writes, tests and ships production code in any language on its own cloud computer, with databases, sign-in, background jobs, domains and hosting. It connects to 1,500+ apps and works inside apps with no API through its own browser. It composes models inside each task, so Gemini handles the steps it is best at and Claude, GPT, Grok, Kimi, DeepSeek or SPIRITT's own Kay models handle the rest. It can also run other agents, such as Claude Code or Codex, in the same workspace. Every action follows allow, ask or never rules per app, with budgets and a full audit trail.
A worked example: from AI Studio prototype to running system
A founder prototypes a customer portal in AI Studio Build mode in an afternoon: login with Google, an orders list, a file upload. She syncs it to GitHub. Then she asks SPIRITT: "Take this repo, move the data to our Postgres, connect orders from Shopify, add a nightly sync, deploy it on our domain, email me a daily summary of new uploads, and alert me if the sync fails." SPIRITT reads the code, writes the integration and tests, deploys it, sets the schedule and keeps watching it. The prototype came from Google AI Studio; the running system came from the agent.
FAQ
Is Google AI Studio good for coding?
Yes. Google AI Studio is one of the fastest free ways to prototype apps in 2026: Build mode generates full-stack web apps and native Android apps from a prompt, with a live preview, GitHub sync and Cloud Run publishing. For large existing codebases, use an editor-based agent such as Antigravity or Gemini CLI, or an autonomous agent such as SPIRITT.
Is Google AI Studio free for coding?
The studio is free with a Google account, and Build mode runs on free-tier Gemini models within rate limits. You pay for paid-tier API usage, for models such as Gemini 4 Carbon that are not on the free tier, and for Cloud Run hosting beyond the Starter Tier's two free apps.
What is Google AI Studio Build mode?
Build mode is AI Studio's vibe coding workspace. You describe an app and the Antigravity Agent generates it as a React and Node.js web app or a Kotlin Android app, with Firebase sign-in and database setup, a live preview, code editing and publishing.
Which Gemini model is best for coding in Google AI Studio?
Gemini 4 Carbon is Google's most capable model for complex software engineering, on the paid tier only. Gemini 4 Hydrogen and Gemini 3.8 Flash have free tiers and suit everyday coding and agents.
Google AI Studio vs Antigravity: which should developers use?
Use Google AI Studio to prototype in the browser and get API keys. Use Antigravity, Google's agent-first desktop app with a CLI and IDE extensions, when you code in your own repository every day.
Can Google AI Studio deploy my app?
Yes. Apps publish to Cloud Run with a shareable URL, and you can connect a custom domain. Eligible personal accounts can publish two apps without a billing account; more need a paid Google Cloud project.
What is the difference between Google AI Studio and SPIRITT?
Google AI Studio builds and publishes apps with Gemini models. SPIRITT is an AI agent that builds, deploys and operates apps and runs the business work around them across 1,500+ apps, composing Gemini with other leading models inside each task, on its own cloud computer with approvals and budgets.
How we checked
SPIRITT publishes this guide. Google AI Studio features, Build mode, publishing options and Gemini API prices come from ai.google.dev documentation (Build mode, publishing and pricing pages), read on November 22, 2026; Antigravity plans come from antigravity.google and Google's AI plans page as reported in November 2026; the Jules wind-down date comes from Google's Jules site. Lovable and Replit prices come from their pricing pages, checked in November 2026. SPIRITT capabilities and prices come from spiritt.ai and the live SPIRITT plan catalog. We read the top Google results for "Google AI Studio for coding" on November 22, 2026. The founder example is a composite, not a real company.