agents-cli
Google's CLI and skill set that helps a coding agent build, evaluate and deploy agents on Google Cloud with ADK and Gemini Enterprise
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Deploys agents to Google Cloud, provisions infrastructure and handles secrets
- Runs build, test and deploy commands in the project
Install
Manual install
uvx google-agents-cli setupInstalls the CLI and skills into detected coding agents. Requires Python 3.11+, uv and Node.js.
This is third-party code. Review the repository files before installing.
What it does
A package built around the agents-cli command line and seven skills that teach a coding agent the full lifecycle of building agents on Google ADK. The CLI creates a project (create, scaffold enhance, upgrade), runs and lints it locally, runs evaluation (generate, grade, compare, analyze, optimize, dataset synthesis, LLM-as-judge), deploys to Google Cloud (Agent Runtime, Cloud Run, GKE), sets up CI/CD and registers the agent with Gemini Enterprise. The skills give the agent a reference for the ADK API, evaluation practices, deployment and observability via Cloud Trace and logging. It works on top of any coding agent, among them Claude Code, Codex and Antigravity. Local development needs only an AI Studio key; cloud features need a Google Cloud account.
Who it is for. For developers building and deploying agents on Google ADK and Google Cloud.
Good fit when
- You need to create an ADK agent project from scratch and run and evaluate it locally
- You need to deploy an agent to Google Cloud (Agent Runtime, Cloud Run, GKE) and set up CI/CD
- You need to register an agent with Gemini Enterprise
- You want one set of skills so the coding agent knows ADK commands and practices
Not a fit when
- Your agent is not on ADK and not tied to Google Cloud
- You need the agent framework itself rather than a wrapper around it, which is ADK
Example request
Use agents-cli to build an ADK agent that compresses long text into short technical notes, and run the evalLimitations
Requires Python 3.11+, uv and Node.js. Local development (create, run, eval) works with an AI Studio key, but deployment and cloud features require a Google Cloud account. AI Studio, the Gemini API and Google Cloud are not reachable from Russia without a VPN, so most features need workaround access.
How to disable. Remove the installed skills from the agent's skills directory, and the plugin in Claude Code via /plugin. The package itself is removed like any CLI, for example uv tool uninstall google-agents-cli or pip uninstall google-agents-cli.
Security check
- Deploys agents to Google Cloud, provisions infrastructure and handles secrets
- Runs build, test and deploy commands in the project
README in short
The README describes agents-cli as a CLI and skill set that turns a regular coding agent into an expert at building, evaluating and deploying ADK agents on Google Cloud. Installation is uvx google-agents-cli setup, and the skills alone can be added via npx skills add. It ships seven skills: workflow, ADK API, scaffolding, evaluation, deployment, Gemini Enterprise publishing and observability. Command tables cover login, project creation, run, lint, the full eval lifecycle, deployment and infrastructure provisioning. Apache-2.0 licensed.
FAQ
Is it a replacement for Claude Code or Codex?
No. It is a tool for a coding agent, not an agent itself: it provides the commands and skills that make the agent better at building, evaluating and deploying ADK agents.
Do I need Google Cloud?
Local development works with an AI Studio key. Deployment and cloud features need a Google Cloud account.
Related
Open-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Open prompt library with a Claude Code plugin, MCP server and CLI: search, fetch and improve prompts and skills from an agent