Google’s Android team released a new command-line interface built specifically for AI coding agents, alongside a skills repository and a knowledge base that gives agents access to up-to-date official documentation. In internal experiments, the Android CLI reduced LLM token usage by more than 70% and completed development tasks 3x faster than agents navigating standard toolsets, according to the Android Developers Blog.

The tools are agent-agnostic. Google explicitly designed Android CLI to work with Claude Code, OpenAI Codex, Cursor, Gemini CLI, Antigravity, or any other agent a developer prefers.

What Android CLI Does

The core problem Android CLI solves: AI agents attempting to build Android apps through standard toolsets burn enormous context tokens trying to understand and navigate Android Studio’s full environment. The CLI strips that down to a lightweight, programmatic interface with direct commands for the workflows agents need most.

Key capabilities include SDK management (android sdk install for specific components), project creation from official templates (android create), virtual device creation and management (android emulator), and deployment (android run). The android update command keeps the CLI current, as reported by SiliconAngle.

Skills and Knowledge Base

Google also launched the Android Skills GitHub Repository, a collection of modular, markdown-based instruction sets (SKILL.md files) that trigger automatically when a prompt matches specific metadata. The initial release includes skills for Navigation 3 setup and migration, Android Gradle Plugin 9, XML-to-Compose migration, R8 config analysis, and edge-to-edge support.

The third component is the Android Knowledge Base, accessible through the android docs command and already available in the latest version of Android Studio. It gives agents access to frequently updated content from Android developer docs, Firebase, Google Developers, and Kotlin docs, solving the training cutoff problem that leads agents to produce code based on outdated patterns.

The Token Economics

The 70% token reduction has direct cost implications. At current API pricing, a 70% reduction means agent-driven Android development costs roughly one-third as much per task. The 3x speed improvement compounds that: fewer tokens per task, and each task finishes in a third of the time.

Google did not disclose which specific agents or models were used in its internal benchmarks, or the exact task distribution tested. SiliconAngle noted that Google “did not explain exactly where these optimizations came from,” though the architecture implies the gains come from replacing unstructured GUI navigation with direct CLI commands that agents can execute in fewer steps.

Why Google Built It Agent-First

The “any agent” framing is a deliberate positioning choice. Rather than locking Android development tooling to Gemini, Google is treating the CLI as neutral developer infrastructure. The strategy tracks with Google’s broader approach during a week where Android CLI, the Android Skills repo, and the Knowledge Base all launched together as a coordinated suite.

Android CLI is available in preview today at d.android.com/tools/agents.