Google on Tuesday announced a new agent for its Opal application designed to create automated workflows. The feature allows users to build mini-apps capable of planning and executing tasks through text prompts. This update integrates the Gemini 3 Flash model to handle the underlying logic and automatically select necessary tools for task execution. The agent operates autonomously to plan subsequent steps based on user input and defined objectives.
The new agent utilizes Google Sheets to maintain memory across different sessions, ensuring data persistence for applications such as shopping lists. Google states this functionality enables users without technical expertise to construct complex workflows within their applications. The agents are designed to be natively interactive, prompting users for additional information or presenting choices to determine next steps when required. This conversational capability facilitates the creation of dynamic, responsive mini-apps tailored to specific user needs.
Google originally launched Opal for users in the United States in July 2025, introducing a platform for creating and remixing mini web apps. By October 2025, the company expanded access to 15 additional markets, including Canada, India, Japan, South Korea, Vietnam, Indonesia, Brazil, and Singapore. A subsequent rollout made the tool available in over 160 countries the following month. In December 2025, Google integrated Opal into the Gemini web app, introducing a visual editor that requires no coding to build custom applications.
The market for natural-language app creation tools is seeing increased activity from various startups. Competitors in this space include Lovable and Replit. Other emerging companies are Wabi, founded by the former head of Replika, and Emergent, which is backed by SoftBank and Lightspeed. Additionally, Rocket.new, supported by Accel, is gaining prominence as a provider of similar development platforms.
