Google kicked off this year’s I/O keynote with a clear message for developers: the Android ecosystem is shifting into an era defined by agents, adaptive design, and AI‑powered workflows. If last week’s early announcements set the stage, today’s keynote filled in the details, showing how Google plans to turn Android into a platform where intelligent tools aren’t just add‑ons but core building blocks. The company even acknowledged that many of the most anticipated updates had already been teased during The Android Show: I/O Edition, which made today feel like the moment everything snapped into place.
The keynote opened with a focus on agentic development, and Google wasted no time highlighting Android CLI, now officially stable. The tool gives AI agents like Claude Code, Codex, and Google’s own Antigravity the ability to perform real Android development tasks, from semantic symbol resolution to Jetpack Compose preview rendering. Google framed this as a bridge between lightweight agent workflows and the heavy lifting of Android Studio, a pairing meant to help developers move faster without sacrificing polish. It’s also the foundation for Journeys, a new system that lets agents run end‑to‑end UI tests under developer direction.
From there, the keynote shifted to Google AI Studio, which now allows developers to generate full Android apps from a prompt. These apps follow modern best practices like Jetpack Compose and Kotlin, and they can be tested through an embedded emulator before being exported to Android Studio for deeper refinement. Google positioned this as a way to prototype quickly while still giving developers the option to take their work all the way to production. It’s a continuation of the theme that’s been building over the past week: AI isn’t just assisting development, it’s becoming a first‑class participant in it.
AI also took center stage with Android Bench, Google’s LLM leaderboard for Android development challenges. The company announced that it’s adding popular open‑weight models like Gemma 4 to help developers compare performance and choose the right model for their workflows. And for teams looking to bring existing apps to Android, the new Migration Assistant in Android Studio can now port iOS, React Native, and web apps into native Android projects. Google showed how the tool can map features, convert assets, and apply Jetpack Compose patterns in hours instead of weeks, a capability it previewed during the keynote itself.
The keynote then moved into the future of intelligent apps. Google introduced Gemini Nano 4 for on‑device tasks like summarization and data extraction, along with Firebase AI Logic for cloud‑based grounding using Maps, URLs, and web search. The new Agent Development Kit and communication protocols like AG‑UI and A2UI are designed to help developers build autonomous, agentic experiences that feel native to Android. AppFunctions, now available for experimentation, lets apps act as on‑device MCP servers, contributing functions that agents can call. It’s a technical shift, but one that fits neatly into Google’s broader push toward agent‑driven app behavior.
A major theme this year is adaptability. Google formally declared Android a Compose‑first platform, with Views now in maintenance mode. Jetpack Compose continues to evolve with new layout options, improved transitions, and better input support. The company emphasized that adaptive design is no longer optional, especially with more than 580 million large‑screen devices in use and users on multiple devices spending significantly more on apps. Notability’s Android debut served as a showcase for what’s possible when apps embrace adaptive principles from the start.
Googlebook, the new high‑performance laptop announced last week, also made an appearance. Google encouraged developers to test their apps on the new Desktop Emulator in Android Studio Canary to ensure they scale gracefully to desktop‑class environments. Jetpack Glance is evolving into a unified widget framework across mobile, Wear OS, and cars, powered by RemoteCompose for high‑fidelity animations and efficient rendering on low‑power hardware. It’s all part of Google’s effort to make Android feel coherent across every screen size and device type.
The keynote also highlighted updates for Android Auto and Android Automotive OS, including new templated media app components and expanded parked experiences like immersive video playback. Developers can apply for early access to bring their video apps to these categories. On the XR front, Google announced Developer Preview 4 of the Android XR SDK, with core libraries moving to Beta soon and new hardware access through the Android XR Developer Catalyst Program. These updates are meant to help developers build richer, more stable XR experiences across the ecosystem.
Media creation got a boost with Android 17’s professional‑grade toolkit. CameraXViewfinder simplifies high‑fidelity capture across foldables and tablets, while the Media3 AI Effects library brings features like Magic Eraser and Studio Sound into a single interface. CodecDB offers chipset‑specific encoding recommendations, and ExoPlayer’s new Scrubbing Mode enables smoother seeking. Google framed these updates as a way to reduce development overhead while raising the quality bar for media apps.
Google TV also saw meaningful updates. Pointer remotes are coming to future devices, and developers can start declaring support now. The Engage SDK, formerly the Video Discovery API, improves recommendations, entitlements, and resumption across all form factors. Google noted that the legacy Watch Next API will lose support in the second half of 2027, giving developers a clear timeline for migration.
Performance rounded out the developer‑focused updates. Android 17 introduces tools like the R8 Configuration Analyzer, ProfilingManager, and integrated LeakCanary to help developers manage memory and optimize performance. The new Android Performance Analyzer uses AI to interpret complex traces and generate SQL queries that pinpoint bottlenecks. It’s a technical but important part of Google’s push to make Android apps smoother and more stable across the board.
Google Play also received a spotlight. Play Shorts, a new short‑form video format, is coming to the Play Store to improve app discovery. Gemini will now help users find apps through suggestions in the Gemini app on Android and the web. And Play Console is gaining agentic catalog management and AI‑powered listing generation to simplify global localization. These updates are aimed at helping developers reach more users with less manual work.
The keynote closed with Android 17 itself. The release includes system‑level improvements like a lock‑free MessageQueue, more frequent and less intensive GC cycles, a new contact picker, and an eyedropper API that reduces the need for sensitive permissions. Google also highlighted behavior changes developers need to prepare for, including mandatory large‑screen resizability and restricted local network access. The third beta of Android 17 QPR1 is also out, bringing a minor SDK update for features that couldn’t wait for QPR2. It was a fitting end to a keynote that balanced immediate tools with long‑term platform direction.
