Google has added an agent step to its Opal software for constructing AI-powered mini-apps. Powered by the Gemini 3 Flash mannequin, the brand new agent in Opal allows autonomous workflows that plan, purpose, and execute on the consumer’s behalf, Google mentioned.
Launched February 24 and obtainable to all Opal customers, the agent step upgrades Opal workflows from static mannequin calls to agentic intelligence, in response to Google. Now, as an alternative of manually selecting a mannequin, builders can choose an agent within the “generate” step. The agent then triggers the appropriate instruments and fashions, resembling Internet Seek for analysis or Veo for video, wanted to perform the consumer’s meant mission. The agent may make use of persistent reminiscence, dynamic routing, and interactive chat with the consumer.
With persistent reminiscence, the agent can use Google Sheets to recollect data throughout classes, resembling model preferences or ongoing lists, thus making mini-apps smarter the extra they’re used. With dynamic routing, the agent evaluates work and decides which steps to set off subsequent, bringing autonomy to mini-apps. With interactive chat, the agent can provoke a chat with the consumer to collect lacking data or provide selections earlier than shifting to a plan’s subsequent stage.
