At the annual Android Dev Summit, Google showcased a transformative evolution in its developer ecosystem, introducing a wave of tooling updates designed to bridge the gap between human engineering and autonomous AI agents. Whether working locally within Android Studio or executing commands from the terminal via the Android CLI, the core mission remains clear: equipping developers and their AI agents with the deep knowledge and sophisticated infrastructure required to build high-quality applications faster and with greater confidence. The summit’s major announcements centered on three primary pillars: unifying Android Studio with any agent and model of choice, expanding the capabilities of the Android CLI to build and deploy from anywhere, and introducing advanced tools designed to ship higher-performing apps with robust reliability. Read Also: Google Expands Access to Advanced AI Workflows with New Linux Terminal Preview on Googlebook Google Play Expands "More Than a Store" Vision with New Discovery Tools, Engagement Features, and Developer Rewards Any Agent, Any Model, and Android Studio Android Studio has long served as the premier environment for professional developers crafting experiences for the Android ecosystem. With the latest updates, the platform is expanding to become an open, flexible workspace optimized for whichever AI agents and models development teams prefer to utilize. A central announcement addressing this flexibility is the stable rollout of Bring Your Own Agent (BYOA) in Android Studio, anchored by today’s release of Android Studio Rabbit 2. Responding directly to developer feedback for interoperability with pre-existing team workflows, Android Studio now natively supports over 40 ACP-compatible coding agents through the open Agent Client Protocol. By navigating to the tool settings, developers can easily browse, install, and seamlessly switch between multiple agents in a unified marketplace. Crucially, these integrated agents gain context-aware access to all the fundamental features that make Android Studio robust, including deep project and build structures, symbol resolution, Compose Preview rendering, the Emulator, and Logcat. BYOA also introduces direct integration with Google Antigravity, ensuring immediate access to cutting-edge Gemini models and upcoming Antigravity features upon launch. Navigating and verifying massive updates generated by AI assistants has historically been a friction point for engineering teams. To mitigate this, Android Studio now incorporates an advanced agentic code review experience called Change Review. Developers can initiate a review session directly from pending agent modifications or the Git changes window to generate an automated report identifying potential bugs and structural concerns. By issuing a simple command within the agent tool window, users can request an AI-guided walkthrough of extensive changes and task the agent with resolving any uncovered issues autonomously. To help teams determine which models and agent harnesses are best suited for particular workloads, Google has introduced Android Bench 2.0. Functioning as a specialized LLM leaderboard, Android Bench evaluates how different models handle long-running, complex real-world Android engineering tasks, such as complete UI migrations or application development from mockups. Unlike standard syntactical checks, these evaluations measure real-world functionality, cost efficiency, and time consumption. Hardware virtualization has also expanded with the stabilization of Android XR emulators directly within Android Studio. Developers can now build, iterate, and test immersive experiences for XR glasses, audio wearables, and display eyewear using customizable 360-degree photographic environments or connected camera feeds. Enhanced with hardware LED indicators for microphone and camera states alongside a streamlined pairing wizard, the updated emulator simplifies virtual and physical device integration. Furthermore, transitioning legacy codebases from XML Views to Jetpack Compose is receiving substantial automation support through the new Migration Assistant. Accessible via right-click contextual menus or project-wide action palettes, the integrated agent collaborates with the developer to draft conversion plans, safely translating layouts, view-binding code, and underlying logic into idiomatic Compose. Building and Deploying Anywhere with Android CLI Complementing the desktop IDE environment, the Android CLI provides a tool-first interface that equips agents and continuous integration pipelines with official Android expertise. By operating directly inside standard terminal environments and CI pipelines, agents can independently build, run, and debug applications without exhausting tokens through trial-and-error reasoning. Testing on physical hardware has traditionally required local physical tethering. However, the latest iteration of the Android CLI introduces remote physical device streaming. Developers and automated agents can provision real physical devices from secure Google device labs, install software packages, and run validation tests identical to a local connection over standard debugging bridges. This is complemented by the Android Device Hub, a lightweight interface providing visualization and interaction controls for local machines, emulators, and remote hardware without necessitating an integrated development environment. To accelerate visual verification cycles, the Android CLI now features native terminal rendering for Jetpack Compose previews via a dedicated command. This allows agents to evaluate graphical layouts across varying screen dimensions independently of an emulator or IDE session. Internal benchmarking indicates that utilizing terminal-based previews dramatically accelerates UI task completion while drastically minimizing token consumption. To streamline agent ecosystems further, Google has released official plugins tailored for major external coding frameworks. The official Android plugin for Google Antigravity brings together CLI tools, official skills, and platform knowledge alongside interactive conversational tools. Similarly, newly released official agent plugins for Claude Code and Codex from OpenAI equip third-party coding agents with deep operational capabilities, including the creation of new Android projects, virtual device management, live UI inspection, journey execution, and targeted knowledge base searches. Shipping High-Performance, Reliable Apps Delivering rapid iteration cycles is only half the battle; users ultimately demand software characterized by elite performance and stability. The summit introduced cross-platform tooling bridging Android Studio and the Android CLI to help engineers transition smoothly from production telemetry signals to verified production fixes. Memory leak detection has been significantly enhanced via a direct integration with LeakCanary within the Android CLI. By performing heap dump analyses locally on the host machine rather than resource-constrained target devices, profiling efficiency is maximized. During demonstrations, capable AI models successfully executed parallel user journeys across multiple simulated devices, leveraged the memory profiling subcommand to isolate reference chains, patched the underlying architectural flaws, and subsequently verified the corrections without impacting runtime performance. A corresponding "Fix with AI" feature is similarly integrated into the Android Studio Profiler. Performance optimization is further augmented by the Android Performance Analyzer, which now underpins terminal-based profiling commands in the Android CLI. Paired with updated profiling skills and the androidx.tracing 2.0 library, systems can capture precise trace metrics detailing source files, specific line numbers, and explicit argument parameters. Addressing production crashes and ANRs that frequently evade local reproduction, the new command-line telemetry integration surfaces vital crashlytics and vitals data directly inside the terminal workflow. This capability works in tandem with modernized code shrinking utilities via R8 optimizations, advanced Jetpack Compose recomposition tracing tools, and a preview of Lightbuild—an early-stage declarative build system structured around human- and agent-readable YAML schemas designed for reproducible, predictable compilation cycles. To ensure models remain synchronized with fast-moving platform paradigms, Google continues to maintain and expand its official Android skills library. Spanning over 30 distinct categories, these modular instructions target specialized domains such as advanced camera architecture, Compose performance tuning, strict security guidelines, and pre-submission Google Play policy auditing. By grounding frontier models with accurate structural guidelines, developers can systematically mitigate compliance risks and accelerate their deployment pipelines. Post navigation Android Studio Rabbit 2 Launches With Bring Your Own Agent Support, XR Emulators, and Advanced AI Tools