Earlier this year, we said we deliver computer usage skills to Advancers in Gemini API. Today, we deliver the model for use Gemini 2.5, our new special model built in Gemini 2.5 The visual proximity of the pre-leading method on the top web and mobile management. Developers can access these skills with Gemini API on Google Ai Studio and Vertex AI.
While AI models can contact the software with organized apis, many digital functions still need direct partnerships with user-clicking, for example, to fill out forms. To complete these tasks, agents should navigate Web pages and apps as people do: click, typing and scroll. The ability to complete forms by birth, monitor applicable activities such as pull and sorting, and work after logs is the next step to build strong agents, powerful agents.
How does this work
The power of the Comol model are displayed with a new `pokt_use tool in Gemini API and should work inside the loop. Input tool is user application, a screen screenshot, as well as the history of the latest actions. Input it can specify and not include unemployment from the complete list of UI actions or specify additional custom entering functions.
