Google plans 'Gemini Spark' agent layer to run agents inside Gemini

Google appears to be preparing a more agentic version of Gemini that can act inside other apps, according to pre-I/O reporting and recent company announcements about Gemini features across Android and Chrome. A 9to5Google report published May 14 said Google is branding that layer “Gemini Spark” and placing it inside the Gemini app. Google has not publicly announced a product by that name. The reporting lands days before Google I/O 2026, which starts May 19, when the company is expected to outline its next wave of Gemini features. Google has already described a broader push toward “Gemini Intelligence” on Android and more automation in Chrome. In a May 12 Android Developers blog post, Google said Gemini Intelligence features would “roll out in waves” starting on recent Samsung Galaxy and Google Pixel phones this summer and later across watches, cars, glasses and laptops. In separate Chrome posts, Google said Gemini-powered browsing tools and “auto browse” are expanding, including on Android next month for U.S. users on Android 12 or higher with at least 4GB of RAM. (9to5google.com) ### Where does “Gemini Spark” fit in Google’s current Gemini plans? 9to5Google reported on May 14 that Gemini Spark is an upcoming AI agent inside the Gemini app and described it as a more advanced agent capability Google has been building in the run-up to I/O 2026. The outlet said onboarding screens point to actions such as decluttering an inbox, preparing meeting briefs and building custom news digests. Because Google has not announced the product, those details remain based on pre-release reporting rather than a public company launch. (developer.android.com) Google’s own public language has been broader and less product-specific. The Android Developers blog said Android is moving toward an “intelligence system,” and the company framed Gemini Intelligence as a suite of features that will arrive in stages rather than all at once. That staged language matches the picture in the pre-I/O report of an agent layer that expands app coverage over time. (9to5google.com) ### Which apps are tied to the reported rollout? 9to5Google said the initial app list tied to Gemini Spark includes Google Tasks, Google Keep and Chrome. The report said Google is planning a gradual rollout, with some automation features not expected until later in 2026. Google has already laid some groundwork for that app mix in public products. (developer.android.com) A Google Workspace post published last year said the Gemini app can connect with Gmail, Drive, Calendar, Keep and Tasks for users on Workspace accounts. Chrome, meanwhile, has become a major surface for Gemini actions, with Google saying the browser can summarize content, manage tasks and automate certain errands with user confirmation for sensitive steps. (9to5google.com) ### What has Google already said about automation in Chrome? Google said in a Chrome blog post that “auto browse” is designed to handle errands such as booking parking or updating orders, while pausing for confirmation on some actions such as purchases or social posts. The company also said deeper integrations across Google apps are coming in the months ahead to support more complex workflows. (blog.google) A separate 9to5Google report published May 12 said Chrome for Android is gaining deeper Gemini integration after Google introduced auto browse on desktop Chrome in January. That matters because Chrome is one of the apps named in the Gemini Spark reporting, and it gives Google a live product surface for agent-style actions before any broader Gemini app announcement. (blog.google) ### How does this connect to Android’s wider AI push? Google said on May 12 that Gemini Intelligence will start on flagship Samsung Galaxy and Google Pixel phones this summer. The Android Developers blog said those features will later spread across other Android device categories, including watches, cars, glasses and laptops. (9to5google.com) Android’s developer materials also point to new plumbing for app integrations. The Android AI site says AppFunctions is an Android API meant to simplify building Android MCP integrations, while Google’s broader AI-on-Android materials describe agentic tools and Gemini-powered experiences across the platform. Those public building blocks suggest Google is preparing both consumer features and developer hooks for more app-aware assistants. (developer.android.com) ### What should readers watch next week? Google I/O 2026 begins May 19, and the conference is the next clear venue for any public confirmation of Gemini Spark or related agent features. Mashable’s preview said Gemini news is expected to be central at the event, while Google’s recent Android and Chrome posts have already set up announcements around Gemini Intelligence and browser automation. (developer.android.com) Late June is the next dated milestone Google has already put on the calendar. Google said Chrome for Android will begin getting Gemini-powered features and auto browse in late June for U.S. users on supported devices, giving the company a near-term rollout point even if broader Gemini app agent features arrive later in 2026. (blog.google) (mashable.com)