Google Labs drops Opal to build AI mini-apps with words

Turning a clever AI idea into an actual working app usually means writing code, wrestling with APIs, or building messy flowcharts on complex low-code platforms. Google wants to skip all that homework.

Google Labs just launched Opal, an experimental tool that lets users describe, build, and share multi-step AI mini-apps using natural language and visual editing. The tool is currently available as a public beta, but only in the US.

Prompt to pipeline without the code

Most real-world tasks require more than a single prompt. You need to chain models, format outputs, and call different tools in sequence to get the job done.

Opal handles this by turning plain English instructions into a complete visual workflow. You tell it what you want the app to do, and it strings together the necessary prompts, model calls, and tools on a visual canvas.

From there, you get full manual control without touching code. You can edit the steps visually, tweak prompts directly, or just type another natural language command to modify the pipeline.

Once built, these mini-apps can be shared via a direct link. Anyone with a Google account can open and run them immediately.

To lower the barrier even further, Opal includes a gallery of starter templates. Users can run these pre-built applications as-is or remix them into custom workflows.

Why it matters

For non-technical creators and operators, Opal looks like a massive shortcut. It eliminates the friction of prototyping simple internal tools or proof-of-concept AI apps. Instead of waiting on developer resources or learning an automation suite, you can talk an idea into an interactive tool in a few minutes.

On the industry side, Google is taking aim at the workflow builder space currently populated by tools like Dify, Zapier Central, and various open-source agent frameworks. By packaging chain-of-thought workflows directly into an accessible consumer interface, Google is trying to keep users building inside its own ecosystem.

Combining conversational prompting with a visual node editor is a smart design choice. Text prompts are great for generating the skeleton, but visual tweaks are much faster for fine-tuning specific steps. The catch right now is its limited scope: US-only, experimental, and geared strictly toward lightweight mini-apps. But for fast prototyping, it is definitely worth keeping an eye on.