Google just shipped a terminal tool built for AI agents doing Android development. Android CLI, now available in preview at d.android.com/tools/agents, gives LLMs a programmatic way to interact with the Android SDK, handling project creation, SDK management, and device deployment from the command line. Internal testing showed it reduced LLM token usage by more than 70% and completed tasks 3x faster than when agents used standard toolsets alone, according to Adarsh Fernando and Esteban de la Canal at Google.
The tool works with Gemini, Claude Code, and OpenAI's Codex. Google is positioning Android CLI as a model-agnostic layer, not just a Gemini feature. The CLI pairs with two other releases: the Android Skills GitHub repository, which houses modular markdown instruction sets that trigger automatically when a prompt matches the skill's metadata, and the Android Knowledge Base, a searchable data source that lets agents pull the latest developer guidelines even if their training data is outdated.
The Skills repository sits under Google's centralized governance. Expert reviewers keep the instruction sets aligned with current best practices. This controlled approach beats letting agents hallucinate their way through outdated Stack Overflow answers, but it also means Google acts as gatekeeper for what counts as correct Android development patterns.
The opt-out telemetry has already raised eyebrows. The CLI collects usage metrics by default, including commands, sub-commands, and flags. You can disable it with the --no-metrics flag, but the default-on approach has developers questioning the choice. And beyond the telemetry concerns, the long-term maintenance burden of apps built heavily with AI-generated code remains an open question, as do the security implications of such assistance.