Blog

Mastering the Agents CLI: Your Gateway to Enterprise-Grade AI Agents

Learn how to build, evaluate, and deploy enterprise-ready AI agents using the Google Agents CLI (ADK).

Posted on: 2026-04-23 by AI Assistant

Mastering the Agents CLI: Your Gateway to Enterprise-Grade AI Agents

The landscape of AI development is rapidly shifting from standalone Large Language Models (LLMs) to autonomous agents—intelligent systems capable of using tools, making decisions, and collaborating to solve complex tasks. To streamline this transition, Google has introduced the agents-cli, a powerful toolkit designed to empower developers and coding agents alike to build, evaluate, and deploy enterprise-grade agents on the Gemini Enterprise Agent Platform.

In this post, we’ll explore the core features of the agents-cli, how to get started, and why it’s a game-changer for the agentic era.

What is agents-cli?

The agents-cli is the command-line interface for the Google Agent Development Kit (ADK). It acts as the backbone of the agent development lifecycle, providing a unified set of tools for everything from initial project setup to production-ready deployment. Whether you are a human developer or an AI assistant like Gemini CLI or Claude Code, agents-cli provides the structure and automation needed to ensure your agents are robust, secure, and scalable.

Key Features

  1. Project Scaffolding (scaffold): Quickly jumpstart your agent projects with pre-configured templates. This ensures your project follows best practices from day one.
  2. Robust Evaluation (eval run): Testing AI agents is notoriously difficult. The eval command allows you to run comprehensive test suites, using LLM-as-a-judge patterns to score trajectories and tool calls.
  3. Seamless Deployment (deploy): Move from local development to production with ease. The CLI integrates with Google Cloud to deploy your agents as scalable services.
  4. Multi-Model Support: While optimized for Gemini 3, the ADK is designed to be flexible, supporting various orchestration patterns and model integrations.
  5. Agent-Centric Workflow: Designed to be used by agents as well as for building them. It includes hooks and metadata that allow coding assistants to autonomously improve and maintain agent code.

Getting Started

Installation is simple, thanks to the uv tool, which handles Python dependencies efficiently.

Prerequisites

Installation

You can set up the CLI using the following command:

uvx google-agents-cli setup

This command will guide you through the initial configuration, including setting up your Google Cloud project and authenticating with the Gemini API.

Core Commands You Should Know

1. scaffold create

Build a new agent from scratch with a single command. Choose from various templates like “search-agent” or “tool-use-agent.”

agents-cli scaffold create --template search-agent my-new-agent

2. eval run

Run your evaluation sets to ensure your agent is performing as expected. You can track scores for accuracy, safety, and efficiency.

agents-cli eval run --evalset tests/eval_suite.yaml

3. deploy

Ready for the world? Deploy your agent to Google Cloud Run or GKE with a single command.

agents-cli deploy --project my-gcp-project

Why It Matters for Enterprise

Enterprise-grade AI requires more than just a clever prompt. It requires traceability, security, and performance monitoring. The agents-cli brings “AgentOps” to the forefront, integrating with Cloud Trace and logging systems to provide deep visibility into how your agents are behaving in the wild.

By standardizing the agent development workflow, Google is making it easier for organizations to build “digital colleagues” that can handle real-world business processes safely and reliably.

Conclusion

The agents-cli is more than just a tool; it’s a foundational piece of the next-generation developer experience. By automating the “boring stuff” of agent development, it allows developers to focus on what truly matters: designing intelligent, helpful, and innovative AI systems.

Are you ready to build your first enterprise agent? Head over to the Google ADK GitHub repository and get started today!

google-adk ai-agents enterprise-ai devops