Google Integrates N8N-like Agent Tool Opal into Gemini Platform

Google has integrated Opal, an agent-building tool, into its Gemini platform, enabling users to create web applications with advanced functionalities. The integration allows for direct generation of web applications featuring interfaces, support for diverse inputs such as images and documents, and web-based output results. The system can also leverage various Google models.

For instance, a user created a screen time analysis tool within Gemini. This tool processes uploaded screen time data to generate a web display, analyze usage patterns, produce a visualized poster, offer text analysis, and create an audio blog based on suggestions.

Accessing and Building Gems

To access this feature, users navigate to the "Explore Gems" option in the Gemini sidebar and select "New Gem." The creation interface presents a simple input box where users describe their desired application. For example, a user requested an application to analyze strength training data from a spreadsheet, providing a report and suggestions.

Upon entering the request, Gemini initiates the Gem application build process, typically completing it quickly. The new Gem supports various input formats, including common files, YouTube videos, web operation recordings, and doodles. A test with training data yielded a detailed output, including a data dashboard, tables, and training suggestions.

Advanced Editing and Model Selection

Users can modify the generated applications by typing prompts directly into the interface. For more complex adjustments, the "Open Advanced Editor" option leads to the Opal interface. This interface displays the workflow, including data processing models and input/output cards.

The Opal interface consists of four main sections:

• Yellow: An area for adding cards, such as models or input items.

• Red: A preview interface for the complete application or individual cards.

• Purple: A prompt-based modification area that synchronously updates the workflow cards.

• Green: The primary editing and adjustment area, where cards representing data or model processing steps can be linked.

While simple modifications can be made via prompts, specific effects or model adjustments require clicking on individual cards and fine-tuning settings. For example, selecting a "Generate Suggestions" card allows users to choose different models and adjust prompts.

Opal offers a range of models for various tasks. For text, audio, video, table, and image input processing, options include Gemini 2.5 Flash or Gemini 3 Pro. Image output can utilize Imagen 4 for text-to-image generation, or Nano Banana Pro and Nano Banana for image-to-image editing. Video generation is supported by the Veo model, text-to-speech by AudioLM, and music generation by Lyria 2. Users can override model selections if the system chooses an inappropriate model.

Application Example and Sharing

In a screen time analysis application, the workflow involves an input card for user-uploaded screenshots. The central processing includes material generation, text analysis, image generation for a visualized poster, and the creation of a spoken script for an audio blog using AudioLM. A final green card generates a webpage to display these materials.

Users can share their applications via a "Share App" button, which provides a link. Recipients can use the application by logging in with a Google account, with model quota deductions based on their Gemini membership level. An example screen time analysis tool is available for modification and editing via a "Remix" option.