Google Connected Apps: Can Your App Work Inside Gemini?

Yes, if your app or website lets customers do something: book, order, check a status or update an account. Google's new Connected Apps Platform lets a business expose those actions so people can use them from inside Gemini and AI Mode in Search. The preview is open only to US developers for now. The groundwork is the same everywhere, though, and Indian businesses can do it today.

What did Google announce?

On 28 October, Google opened its Connected Apps Platform in public preview. It is one place for businesses and developers to connect their apps to Gemini and Search. A user links your service to their Google account once. After that, they can ask Gemini to do things through it.

Gemini already has more than 50 of these apps, including Adobe, Experian and Peloton. Search has fewer so far: Instacart, Canva, YouTube Music and Yahoo Fantasy Football. Google says the Gemini apps will reach Search over the next few weeks.

You can see where this is heading in the Instacart example from July. You ask AI Mode for a recipe, it adds the ingredients to your Instacart cart, and you check out on Instacart. Google does the asking. The business still owns the transaction.

How does an app get into Gemini and Search?

According to Google's developer page, there are two technical routes:

1. A cloud-hosted MCP server. MCP (Model Context Protocol) is an open standard for describing what your system can do, such as "check availability" or "create booking", so an AI assistant can call it safely. This route suits most web businesses.
2. Android MCP through AppFunctions. If you have an Android app, it can expose actions that run on the phone itself.

Users sign in through OAuth, so Gemini acts with their permission. You then register the app in Google Connected Console, test it, pick Search, Gemini or both, and submit it for Google's review. Once it's approved, you decide when it goes live. You also get numbers on traffic, latency and errors.

The useful part for founders: OpenAI's apps in ChatGPT, open for submissions since December 2025, are built on the same MCP standard. Build the layer once and most of it carries over to both assistants.

Is it worth it for your business?

It depends on what customers do with you. A brochure site has nothing for an assistant to call.

What should you check before building one?

This is the readiness checklist we'd run before writing any MCP code:

• Clean APIs. Every action you want to expose (search, book, cancel, pay) has a documented server-side endpoint. If the logic lives only in front-end code or a no-code tool, that comes first.
• Real accounts. Customers log in, and the backend checks who they are before it shows or changes anything. An assistant acting for a user must never see another user's data.
• OAuth support. Your login can issue tokens to a third party with limited permissions. Many in-house logins can't do this yet.
• Plain error messages. "Slot no longer available" beats a stack trace. The assistant relays whatever you send back.
• Speed. Slow endpoints fail inside a conversation.
• Rate limits and logs. See what the assistant called, and cap it if something loops.
• A privacy check. If the assistant handles personal data, your privacy notice should say so. That matters more as India's DPDP Rules phase in.

What can Indian founders do while it's US-only?

Google says more markets will open in phases over the coming months, but it hasn't named dates. Use the wait to get the plumbing right. For many apps, the first two checklist items are the real work. Fixing them also makes your website, mobile app and WhatsApp bot easier to build on.

A sensible order is: tidy the API, add proper OAuth, then build one MCP server with two or three high-value actions. Test it with ChatGPT first, since submissions there are already open. Then you're ready to apply to Google the day your market opens.

Don't expose everything. Start with the one action your customers repeat most, such as booking or order status, and measure whether anyone uses it.

Want your app ready for Gemini and ChatGPT?

We build the API, login and MCP layer that lets AI assistants act on your system safely. Our AI and LLM integration service starts with a review of your current setup and ends with a fixed price before any work begins. Tell us what your customers do in your app and we'll reply within 4 working hours with what's needed.