Google debuts no-code Opal tool for building ‘AI mini apps’

Google LLC has introduced an artificial intelligence tool that enables users to build simple applications without writing any code.

The tool, which is known as Opal, made its debut on Thursday. The launch comes a couple of weeks after Amazon Web Services Inc. introduced an IDE, or integrated development environment, called Kiro that uses AI to speed up software projects. 

The starting point of an Opal development project is a ChatGPT-like chatbot interface. Users enter a natural language description of the task their application will perform. They can also provide any additional data that may be needed to complete the task, as well as connect external tools.

The service uses the provided data to automatically generate a simple application. Once the software is ready, users can share it through a button at the top right corner of the interface. It makes the application accessible through a sharing link similar to those used by Google Docs.

Optionally, workers can edit an Opal-generated application before publishing it using a visual editor. The editor represents an application as a collection of cards on a virtual canvas. Some of the cards describe the input the application ingests, while others control how it processes that input.

The visual editor makes it possible to link together multiple data processing steps. For example, a user could have an application generate a product description and then create a promotional video based on that description. Workers can customize how such tasks are performed by entering natural language instructions via Opal’s visual editor. 

“Opal converts your app description into a multi-step workflow with inputs, generation steps and output steps,” Google product manager Elle Zadina explained in a demo video. “You can click into a step to see the prompt or instructions and edit them directly.”

Under the hood, Opal relies on multiple AI models to generate applications. Google didn’t specify which algorithms are used. One candidate is the company’s recently introduced Gemini 2.5 Pro large language model, which is particularly adept at application development tasks. It has set a record on the popular LMArena benchmark for measuring LLMs’ coding skills.

Given that Opal is geared towards creating simple applications, it may not require the advancing programming features offered by Gemini 2.5 Pro. Google may be using a less capable but more power-efficient coding model to power the tool. Opal’s clip generation features, in turn, are presumably based on one of the company’s Veo video algorithms.

Opal is rolling out with more than a half-dozen prebuilt applications. Some focus on consumer use cases such as designing video games. The rest enable workers to generate marketing materials, research products and perform other business tasks.

On launch, Opal is available to users in the U.S. through a public beta program. The tool can be accessed via Google Labs, a webpage that hosts more than two dozen of the search giant’s experimental AI products. They automate tasks ranging from searching the web to generating application interface designs. 

Image: Google