Google expands Gemini across Android

Google on May 12 said it is expanding Gemini from a chatbot and coding assistant into a broader Android layer that can act across apps, screens and devices. The company introduced “Gemini Intelligence” at its Android Show, describing Android as moving from an operating system toward what its developer blog called an “intelligence system.” The first rollout is limited: Google said the features will arrive on select Samsung Galaxy and Google Pixel phones this summer, with wider availability across watches, cars, glasses and laptops later in 2026. ### What is Google actually adding to Android? Google’s product post said Gemini Intelligence is designed to automate multi-step tasks, use screen and image context, summarize web content and help users complete forms. The examples Google gave were specific: booking rides, building shopping carts, finding a class syllabus in Gmail and using that information to add books to a cart. Google also said the system is being tuned on the Galaxy S26 and Pixel 10 with selected food and rideshare apps. (blog.google) Matthew McCullough, Google’s vice president of product management for Android Developer, wrote on May 12 that Gemini’s task automation is expanding across selected apps “on behalf of the user” with transparency and control. His post said the capability first launched with selected food and ridesharing partners and is now extending to more verticals and form factors, including foldables, watches, cars and XR glasses. (blog.google) ### How do developers plug their apps into that system? Google’s Android developer blog said AppFunctions is the main integration path for developers who want more control over how agents interact with their apps. The framework lets developers expose services, data and actions to the operating system and its agents, paired with natural-language descriptions, so the system can discover and execute those tools across devices. (developer.android.com) The same post said Google has started testing the early-stage APIs in a private preview with KakaoTalk for actions such as sending messages and starting voice calls. Google also said AppFunctions has already enabled local execution of 25 app use cases across device manufacturers, though it did not publish a full list in the post. (developer.android.com) ### Where does Gemini CLI fit into this story? Gemini CLI is a separate Google tool aimed at developers working in the terminal, not a consumer Android feature. Its documentation says the CLI is meant to use Gemini models with local project context to understand code, automate tasks and build workflows. The hooks system lets developers run scripts at defined points in the agent loop to inject context, validate actions, enforce policies, log interactions or block risky tool use. (developer.android.com) Gemini CLI’s release notes show Google is still adding core agent features quickly. The stable changelog lists versions 0.41.0 on May 5 and 0.42.0 on May 12, with additions including real-time voice mode, stricter environment loading, shell-command validation, memory tools and compliance warnings. Those updates matter because they show Google building the controls around agent behavior alongside the automation itself. (geminicli.com) ### What about BigQuery and AlloyDB hooks? Gemini CLI’s official documentation does not present BigQuery and AlloyDB as built-in Android features. Instead, Google’s Gemini CLI materials describe extensions as packages that can add MCP servers, prompts, hooks, sub-agents and other capabilities, and GitHub materials tied to the project list BigQuery and AlloyDB among extension examples for querying datasets and interacting with databases. (geminicli.com) That indicates the data-querying use cases sit in the developer tooling layer around Gemini, rather than in the Android consumer rollout announced this week. The practical use case is straightforward. A developer can use Gemini CLI with extensions or hooks to pull structured project context into an agent workflow — for example, querying logs, database records or other operational data — while Android’s Gemini layer handles user-facing tasks on devices. Google’s hooks documentation also says those scripts can review tool arguments, block dangerous operations and log activity, which underlines that the company expects human-defined guardrails around agent actions. (geminicli.com) ### When do the next details arrive? Google said Gemini Intelligence will begin rolling out on select Samsung and Google phones this summer, with broader device coverage later this year. Google I/O 2026 begins online on May 19 at 10 a.m. Pacific time, and the company has said the event will include updates on Android, Gemini, Chrome and Cloud. (blog.google) (geminicli.com)