Google ships Agents‑CLI demo

Google Cloud on April 22 released Agents CLI, an open-source command-line tool meant to help coding assistants build, test and deploy AI agents on Google Cloud. (developers.googleblog.com) (github.com) Agents CLI is not a chatbot itself. Google describes it as a command-line interface plus bundled “skills” that give tools like Gemini CLI, Claude Code, Codex and Cursor machine-readable access to Google Cloud agent workflows. (developers.googleblog.com) (github.com) The project’s GitHub repository went public with an Apache-2.0 license and showed about 1,600 stars and 173 forks when indexed this week. The latest listed release was v0.1.1 on April 22, 2026. (github.com) Google says the tool is built around the “Agent Development Lifecycle,” or the sequence from scaffolding a project to evaluating it, deploying it and monitoring it after launch. The company’s blog says developers can move from concept to a live service in “hours, not weeks.” (developers.googleblog.com) In plain terms, the problem Google is trying to solve is context overload. Instead of forcing a coding assistant to read scattered product docs for Cloud Run, Google Kubernetes Engine, Agent Platform and related services, Agents CLI exposes those steps as commands and packaged skills. (developers.googleblog.com) (github.com) The core commands in the public README include `agents-cli setup`, `agents-cli scaffold`, `agents-cli eval run`, `agents-cli deploy` and `agents-cli publish gemini-enterprise`. Google also lists observability support through Cloud Trace, logging and third-party integrations. (github.com) Google’s launch post uses a sample prompt for a travel-expense agent that auto-approves expenses under $50 and routes larger or unusual items to a human reviewer. The example then scaffolds a project with a command targeting `agent_runtime` and runs evaluations against a ground-truth dataset. (developers.googleblog.com) The deployment targets named in Google’s docs include Agent Runtime, Cloud Run and Google Kubernetes Engine, or GKE. The docs site also presents the tool as part of “Agent Platform,” Google’s broader push to turn its Agent Development Kit into a production stack. (github.com) (google.github.io) That matters because Google introduced the open-source Agent Development Kit at Google Cloud Next 2025 as a framework for building multi-agent applications. Agents CLI moves that earlier framework closer to a full developer workflow, with deployment, evaluation and registration steps exposed in one surface. (developers.googleblog.com 1) (developers.googleblog.com 2) Google’s pitch is that developers should be able to use their preferred coding agent rather than learn every cloud product interface directly. The repo’s README says the package works with “any coding agent,” but Google’s named examples are Gemini CLI, Claude Code, Codex and Antigravity. (github.com) The launch does not, by itself, prove Google’s workflow is faster than rival agent stacks from other vendors. What it does show is Google packaging its agent framework, cloud deployment targets and observability tooling into a single open-source CLI just six days before this report. (developers.googleblog.com) (github.com)