Can Gemini build an app? Yes. Here are Google's routes and what each one costs

Short answer: yes. Gemini can build a working app from a plain-language description. For a small app you share by link, use Canvas in the free Gemini app; for a full web or Android app you can publish, use Build mode in Google AI Studio. What Gemini does not settle for you is where the app runs afterward, so most of this post is about that part.

We build VibeKit, one of the options below. Every claim about Google's tools here was checked against Google's own pages on November 9, 2026, and prices are US prices as of November 2026.

Which Google tool should I use to build an app?

Google has four main ways to build with Gemini right now, and they suit different people:

If you have never written code, start with the first two.

Can the Gemini app build an app for free?

Yes. Canvas is "available for all Gemini users." You choose Canvas in the Gemini app, describe what you want, and Gemini writes the code and shows a live preview beside the chat. You change it by asking, or by editing the code view directly.

Canvas apps are more than mockups. Google's Canvas help page says an app can save data across sessions and devices and share data between users, which covers a family list or a score tracker. Know three things before you share one:

• Anyone with the public link can view and edit the app's data. Google says so in plain words. Fine for a quiz, wrong for anything private.
• Shared links open only at gemini.google.com, not in the Gemini mobile app.
• Adding AI features to an app requires being 18 or over.

Canvas is the right choice when the app is small, the audience is friends or a class, and living at a Gemini link is fine.

Can Google AI Studio build a real app?

Yes, and it is Google's main route for a full app. Build mode turns a prompt into a project with a React front end and a Node.js server, written by Google's Antigravity agent. You refine it in chat, or highlight part of the screen and describe the change.

Per Google's docs, Build mode can:

• Set up a Firestore database and "Sign in with Google" automatically.
• Connect Gmail, Sheets, Docs, Drive and Calendar.
• Publish to Cloud Run, Google's hosting service. "Pricing for Google Cloud Run may apply based on usage," and a custom domain needs a paid Google Cloud project.
• Sync both ways with GitHub, or download a ZIP to run elsewhere.
• Build native Android apps in Kotlin, previewed in a browser emulator and publishable to the Play Store for testing. iPhone apps are not on the list.

One detail matters once other people use your app: a Cloud Run deployment uses your Gemini API key for every visitor's Gemini calls, and Google notes that calls from shared apps count toward your usage limits. AI Studio's pricing page offers a free tier with free tokens and limited model access. The flagship, Gemini 4 Argon, is not on the free tier, and its standard paid price doubles on December 1, 2026: input goes from $2 to $4 per million tokens, output from $10 to $20.

What about Gemini CLI, Antigravity and Jules?

These three are for people comfortable reading code.

• Gemini CLI is no longer the free terminal tool many tutorials describe. Its official repository says free individual access and Google AI Pro and Ultra moved to Antigravity CLI on June 18, 2026. Gemini CLI now serves organizations on Gemini Code Assist Standard or Enterprise.
• Google Antigravity is Google's "agentic development platform": an IDE, a command-line tool and a desktop app, for macOS, Windows and Linux. The individual plan is $0 with basic weekly rate limits, and Google AI Pro and Ultra raise them. It works on your own machine, so hosting is still your decision.
• Jules clones a GitHub repository into a cloud machine, shows you a diff of its changes, and opens a pull request. Per its usage limits, the free plan allows 15 tasks a day, Jules in Pro 100 and Jules in Ultra 300. It improves code you already have; it does not host anything.

Is Firebase Studio or Opal still an option?

Not for new projects, though older guides still recommend both. Google's Firebase Studio sunset page says new workspace creation and signup were disabled on June 22, 2026, and that the service shuts down on March 22, 2027, when remaining data is deleted. Google points people to AI Studio or Antigravity instead. Firebase itself (Firestore, Authentication, App Hosting) is not affected.

Opal, the Google Labs mini-app builder, "is graduating out of Labs on November 17, 2026," and its active workflows will no longer run after that date.

How much does it cost to build an app with Gemini?

Building can cost nothing: Canvas is free, and AI Studio has no cost to start. Google's AI plans raise usage limits. Google AI Plus is $4.99 a month, Google AI Pro is $19.99, and Google AI Ultra starts at $99.99, with a $199.99 tier above it.

Running the app is the recurring part. On Google, Firebase's no-cost Spark plan includes Hosting (10 GB stored, 360 MB a day of transfer) and a Firestore allowance, while App Hosting requires a billing account. Cloud Run bills by usage. Our cost breakdown covers the general picture, and the free-route guide compares the $0 options.

Why doesn't generated code feel like an app yet?

Because writing the code is only the first step. An app other people can open also needs:

1. A server that is running when a visitor arrives, not only while your browser tab is open.
2. A database that keeps data through restarts and updates.
3. A stable URL, and often your own domain.
4. A way to update it without losing what people already saved.
5. Someone paying for the AI calls visitors trigger, with the key kept off the page.

Canvas handles a small version of this inside Gemini. AI Studio handles it by publishing into a Google Cloud project you own, which means you also own the billing account, the key and the bill. That is a good deal if you want Google's stack, and extra work if you just want the app. More on the options in where to host AI apps.

Can an app builder run Gemini for me?

Yes. App platforms run the coding agent and host the result, so the five steps above are handled for you. In VibeKit, Gemini is one of the models in the picker. It runs as a platform model paid from your VibeKit credits, and there is no option to bring your own Google key. Your own key works for Anthropic and OpenAI only (how that works), and once one is connected your agent uses that provider instead. There is also an always-free model tier, but it does not run Gemini.

Each app gets its own coding agent that remembers the project, and a live address at <name>.vibekit.page, hosted by VibeKit on AWS in the US. The Free plan is $0 with 10 AI sessions a month and 2 apps, a database is a $3/month add-on, and AI usage is pay-as-you-go from credits. Free apps sleep after 60 idle minutes and wake on the next visit. Starter is $9.99 a month on the web, with $3 of AI credit, a database included and one always-on app. Full details are on pricing.

Google is the better choice for a tiny shared tool (Canvas is free), a native Android app headed for the Play Store, or an app built around Google sign-in and Workspace data. For iPhone apps, see can AI build an iPhone app. The Claude and ChatGPT versions of this question work the same way.

Try it: one small app, two ways

Start in VibeKit or open the iOS app, create a new web app, and choose Gemini in the model picker. Then paste:

Build a shared grocery list for my household. Anyone with the link can add an item, tick it off, and see who added it. Keep the list saved between visits and give me the live URL.

Open the URL on two phones, add an item on one, reload the other, and check that the item is there. Paste the same prompt into Gemini Canvas as well, and you will see the difference between a Canvas share link and a hosted app for yourself.