This Free Google Tool Lets you Build and Share AI Apps in Minutes - No Coding Required
A hands-on introduction to Google Opal, and how anyone can use it to build simple AI mini-apps
Introduction
Most people still think building apps means writing code, setting up servers, or learning complex tools.
But that’s no longer the full picture.
The software world is changing fast — today, you can build surprisingly useful things with natural language. People call it vibecoding ✨: you describe what you want, the tool fills in the gaps, and you refine until it works.
There are already powerful tools out there (for example, Replit) that can help you generate full apps — but many everyday problems don’t need a “full app.” Sometimes you just need a small mini-app that solves one annoying task you keep repeating (the kind of thing you always end up doing in ChatGPT or Gemini again and again).
That’s where Google Opal comes in.
Opal is a free, no-code tool from Google that lets you build and share AI mini-apps in minutes — either by describing what you want in plain English, or by editing a visual workflow (think: simple connected steps). No servers, no deployment stress. When you’re ready, you can publish and share your mini-app with a link so other people can run it too.
In this article, I’ll break down what Opal is, how it works, and then we’ll build one practical mini-app together — so you can confidently start creating your own.
What Google Opal Lets You Do?
At a high level, Google Opal lets you turn ideas into small, useful AI apps — quickly. If a task can be described, repeated, or automated, there’s a good chance it can be built with Opal.
Instead of writing code, you describe what you want the app to do or connect a few simple steps visually. Opal handles the rest. This makes it easy to experiment, tweak ideas, and build tools that solve very specific problems.
To show what’s possible, the Opal team has already built and shared several example mini-apps. There’s a blog post writer that researches a topic and generates a full article from it. There’s a business profile analyzer that takes a company name and shows how it appears across the internet — from websites to social media and online mentions.
Other examples are more playful but just as practical. A fashion stylist mini-app suggests outfits based on the weather and occasion. A mood-based playlist generator creates a music list with YouTube links just from a short description of how you’re feeling. There’s even a tool that creates a calendar invite from an image of an event — you upload a photo, add a few details, and the invite is generated for you.
People have also built things like personal podcast generators that create audio updates on any topic you choose, and sticker generators that let you design custom stickers beyond the defaults you get in messaging apps.
What all these examples have in common is that they’re not “big apps.” They’re focused mini-apps — tools designed to do one thing well. And that’s the real power of Opal: it lowers the barrier so anyone can build small, useful AI tools and share them with others.
At this point, you’re probably wondering how this actually works. So instead of talking about Opal, let’s build something with it.
The Mini-App We’re Building
For this walkthrough, we’ll build a simple AI mini-app that turns messy records into clean, structured reports.
The idea is straightforward. You upload files — handwritten notes as images, typed documents, or plain text — and the app understands the content, organizes it, and outputs a properly formatted report. Depending on what you choose, the final result can be a clean document or a PDF with tables and clear structure.
This kind of mini-app is useful in everyday situations: financial records, attendance lists, reports, or any information that starts off unstructured and needs to be cleaned up. Instead of manually rewriting everything, the app handles the organization for you.
The goal here isn’t to build something complex. It’s to show how a small, focused AI app can solve a real problem.
Now let’s build it step by step.
Step 1: Open Google Opal
To get started, head over to opal.google and make sure you’re signed in with your Google account.
Once you’re on the page, click the Create New button. This opens Opal’s main workspace — the canvas where you’ll design your mini-app.
Step 2: Understand the Workspace
When the canvas loads, you’ll notice two main modes at the top:
Editor – this is where you build and configure your app
App – this shows a live preview of how the app will look and behave when someone uses it
For now, stay in Editor mode.
You’ll also notice two areas on the screen:
The canvas, where steps are placed and connected
A configuration panel, where each step can be edited
Step 3: Choose the Visual Builder
Opal supports two ways of building:
Describing what you want using natural language
Building visually using connected steps
For this mini-app, we’ll use the visual approach, so we can clearly control each part of the process.
At the top of the canvas, you’ll see several building blocks you can add:
User Input
Generate
Output
Add Assets
We’ll start with User Input.
Step 4: Add the User Input Step
Click on User Input.
A new step appears on the canvas.
This step defines what the user will provide when they run the app.
When you select the step, a configuration panel opens on the side. Here, you can:
Rename the step (we’ll keep it as User Input)
Write a prompt that explains what the user should provide
Adjust advanced settings
Step 5: Configure the User Input
In the prompt area, you’ll describe what the user should upload or type.
For now, you can input this “Got any Messy record or data (Text, images, audio, videos...) dump them here lets create something Clean and Structured”
Scroll down to Advanced Settings and set the following:
Input type:
Any
This allows the app to accept text, images, or documents.Required: Enabled
This ensures the app can’t run without user input.
This flexibility is important, since the mini-app is designed to work with different kinds of records — handwritten notes, typed files, or plain text.
At this point, the user input step is ready.
Step 6: Add the Generate Step
Now that the user input is set up, the next step is telling Opal what to do with that input.
At the top of the canvas, next to User Input, click Generate.
A new Generate step appears on the canvas. It looks similar to the User Input step, but it serves a different purpose — this is where the AI processes and transforms the input.
Click on the Generate step to open its configuration panel.
Renaming and Choosing a Model
Just like the previous step, you can rename this step if you want. The name doesn’t affect how it works — it’s just for clarity. You can leave it as Generate or give it a more descriptive name.
Below the step title, you’ll see a model selector. Opal gives you access to several AI models, including:
Gemini 3 Flash
Gemini 3 Pro
Gemini 2.5 Flash
For most text-based processing tasks, the latest Gemini models work well. You can choose whichever fits your use case.
Opal also supports other types of generation:
Image generation
Text-to-speech
Video generation
For this mini-app, we’ll focus on text generation, since the goal is to clean and structure records.
Understanding Prompts and System Instructions
After selecting the AI model in the Generate step, the next field you’ll see is the Prompt.
This prompt is the user prompt — it’s where you describe what the AI should do with the input it receives.
At this point, you can already start writing instructions like:
What the input represents
What kind of output you expect
How the information should be structured
This prompt works directly with the data flowing in from the User Input step.
If you scroll down and click Advanced settings, an additional field appears: System Instructions.
System instructions define the overall role and behavior of the AI. Think of them as higher-level rules the model should follow consistently, no matter what input it receives. This is where you describe what kind of assistant the AI is and how it should think about the task.
For example, in this mini-app, the system instruction might describe the AI as something like:
A helper that cleans messy records
An assistant that organizes information clearly
A tool that outputs structured, professional reports
You don’t need to overthink this. If you’re unsure how to write good system instructions or prompts, you can use tools like ChatGPT to help. Simply describe the mini-app you want to build and ask it to generate:
A system instruction
A user prompt
You can then copy and paste those directly into Opal.
This is the system Instruction we will be using:
You are an AI that converts messy, unstructured records into clean, structured reports.
You can process handwritten notes (from images), typed documents, videos, audio e.t.c
Your task is to accurately understand the content, organize it logically, and produce a clear, well-formatted report with headings, sections, and tables when appropriate.
Preserve all important information, remove noise or duplication, and improve clarity without adding new facts.
This is the User Prompt :
Task:
Analyze [user_input], extract all relevant information, and reorganize it into a clean, structured report.
Present the information with clear headings, logical sections, and tables where appropriate.
Remove duplicates, fix formatting issues, and improve readability while preserving the original meaning and all important details.
Output:
A professional, well-structured report suitable for viewing as a document or exporting as a PDF.
Connecting the Steps
Once your prompt and system instructions are set, the next step is connecting the workflow.
On the User Input step, you’ll see a small black connector dot at the top-right. Click and drag it onto the Generate step, then release.
When the steps are connected, Opal understands that whatever the user uploads should be passed into the Generate step. You’ll notice that the user input becomes available inside the prompt automatically.
This visual connection is what makes Opal easy to reason about — you can literally see how data flows from one step to the next.
Referencing Other Steps and Tools
Inside the prompt editor, you can also type @ to reference:
Other steps in your workflow
Built-in tools provided by Opal
Additional data sources
When you type @, a list of available options appears, and you can select what you want to include. This makes it possible to build more advanced workflows without writing any code.
Opal also includes an option to suggest edits to your prompt. You can describe what you want to improve, and Opal will help refine it for you.
At this point, the Generate step is fully set up.
The AI knows what role it should play, what input it receives, and how it should process it.
Next, we’ll decide how the final result is shown to the user.
Step 7: Add the Output Step
Once the Generate step is ready, the next step is deciding how the result is delivered to the user.
At the top of the canvas, click Output.
An Output step appears on the canvas. Select it to open its configuration panel.
Just like the other steps, you can rename this step if you want. The name is only for clarity.
Choosing the Output Format
In the configuration panel and you’ll see the output format options. Opal provides several formats:
Manual layout
Displays the content exactly as it’s generated by the AI agent. No additional formatting is applied.Web page (auto layout)
Renders the output as a simple web page using HTML and CSS.Save to Google Docs
Saves the output as a Google Docs document.Save to Google Slides
Creates a presentation similar to PowerPoint.Save to Google Sheets
Structures the output into a spreadsheet.
For this mini-app, select Save to Google Docs, since we want a clean, editable document.
Connecting the Generate Step to the Output
The Output step needs to know where its content is coming from.
You can connect it in two ways:
Drag and connect
Click the small connector dot on the Generate step and drag it onto the Output step.Reference it in the prompt
Inside the Output prompt field, type @, scroll through the list, and select the Generate step.
Both approaches do the same thing — they pass the generated content into the Output step.
Formatting the Final Document
Inside the Output prompt, you can control how the document looks.
This is where you can add static text that always appears in the document, along with the generated content. For example, you might start with a line like:
This is your cleaned and structured report.
Below that text, you insert the output from the Generate step.
This ensures that every document follows the same consistent structure.
Advanced Settings: Document Title
Scroll down and click Advanced settings.
Here, you’ll see an option to define the Google Docs title. This allows you to control the name of the document every time the app runs.
You can set a fixed title, so each document is named clearly and consistently.
At this point, the full workflow is complete:
User Input collects files or text
Generate processes and organizes the content
Output formats and saves the result as a document
Publishing the Mini-App
With the workflow complete, the final step is making the app available to others.
At the top-right corner of the Opal interface, click Share app.
A modal appears with sharing options.
Toggle the app from Private to Public. Once you do this, Opal generates a public link.
Anyone with this link can open and use the mini-app — no setup, no installation, no coding required.
You can now copy the link and share it anywhere.
That’s it.
You’ve built a complete AI mini-app:
Inputs
Processing
Output
Publishing
From here, everything else is iteration and creativity.
I’ve published the mini-app we just built so you can try it yourself.
You can upload your own files, run the workflow, and see how Opal turns messy input into a clean document.
👉 Try the mini-app here:
https://opal.google/app/1yKByyWxN1xHOLcWQnPwafV3gNUUUwKxe?shared
Tools like Google Opal change how we think about building. You no longer need to create big products or learn complex systems to be useful. Sometimes, a small tool that solves one specific problem is enough.
If this mini-app gave you ideas, that’s the point. The real power of Opal is not what I built here — it’s what you can build next.
If you found this useful, feel free to follow me:
X (Twitter): @coollad49
LinkedIn: linkedin.com/in/daniellucasadebayo/
GitHub: github.com/coollad49
More coming soon.