Your coding agent’s field guide to Google Cloud agents

What it does

agents-cli is a meta-tool—a CLI and skill pack that turns coding agents like Claude Code, Codex, or Gemini CLI into specialists for Google Cloud’s agent stack. It covers the full lifecycle, from scaffolding projects and writing ADK Python patterns to running evaluations and deploying to Cloud Run or GKE. You can drive it standalone from the terminal, but it is built to be consumed by an AI coding assistant that does the typing for you.

The interesting bit

Instead of you learning Google Cloud’s deployment and evaluation CLIs, you install these “skills” into your coding agent, which then absorbs lifecycle rules, model selection heuristics, and infrastructure patterns. It is essentially a curriculum and toolchain for agents, wrapped as a CLI.

Key highlights

• Ships as discrete “skills” (e.g., google-agents-cli-deploy, google-agents-cli-eval) that teach coding agents specific domains like ADK orchestration, CI/CD setup, and observability.
• Local development works with an AI Studio API key; deployment targets include Agent Runtime, Cloud Run, and GKE.
• Evaluation suite supports trace generation, LLM-as-judge grading, failure-mode clustering, and prompt auto-tuning.
• Can enhance existing ADK projects with deployment configs, data-ingestion pipelines, and CI/CD.
• Distributed as a pre-built Python wheel rather than source in the repository.

Caveats

• Deployment and cloud features require Google Cloud; local dev is limited to scaffolding and evaluation.
• The project is in a pre-GA preview state with limited support.
• The CLI ships as a pre-built wheel, so the source code is not in the repository—only the release artifact.

Verdict

Worth a look if you are already building ADK agents and want your coding assistant to handle the Google Cloud boilerplate. Skip it if you are not using Google Cloud or ADK; it is tightly coupled to that stack.