Introduction
Managing separate AI agent skills and tools for different Google Cloud tasks gets tedious fast. I have seen developers lose time stitching together individual skills, one for authentication, another for documentation lookup, and yet another for CLI commands. Each piece needs its own configuration and maintenance. The Google Cloud Developer Plugin for AI coding agents bundles these capabilities into a single, installable package. This post covers what the plugin does, how to install it, and how it fits into a real workflow.
Background on Agent Plugins and MCP Servers
AI workflows often rely on two types of tools: agent skills and Model Context Protocol (MCP) servers. Skills are specific capabilities an agent can use, like reading a file or making an API call. MCP servers provide a standard way for agents to access external data and services, such as documentation or live APIs. The problem arises when you need several of these together. An agent analyzing infrastructure is far more effective when it can combine domain knowledge from documentation, workflow recommendations, and the ability to interact with a live environment simultaneously. Managing these connections individually creates a “tool coupling problem” where related capabilities are split across loose configurations.
The open Agent Plugins standard addresses this by packaging skills and MCP servers into portable bundles. Instead of custom wrappers for each AI assistant, developers use a unified manifest and directory structure. This portability means a plugin built for one agent should work with others that support the standard.
What’s Happening Now: The Google Cloud Developer Plugin
The foundational google-cloud-developer plugin focuses on the core tasks of working with Google Cloud as a developer. Its main areas are authentication, project management, and guardrails for gcloud CLI operations. It also bundles configuration for the Developer Knowledge MCP server, which gives agents access to up-to-date Google developer documentation. This means your agent can reference official docs while you work, without manual context switching.
This plugin aligns with the vendor-neutral Agent Plugins specification. The package is hosted in the public Google Agent Skills repository, making it accessible to anyone regardless of their preferred AI coding environment. It is the first in a planned series, starting with these foundational capabilities before moving to more specialized, product-specific plugins.
What It Means in Practice: Installation and Use
Installation is straightforward if your AI agent CLI supports the standard. For the Antigravity CLI, you run a single command with the plugin’s repository URL. For Claude Code or Codex, you first add the Google plugins marketplace, then install the plugin by name. The process is similar to adding a package in a programming language’s ecosystem.
After installation, the plugin provides immediate environmental awareness. Let’s say you need to bootstrap a new project for a script. You could prompt your agent: “I’m new here. Help me get a project set up with billing and authenticate my local machine so my script can call APIs as a service identity.” The agent, now equipped with the plugin, will silently run checks for existing CLI tools and projects, outline a workflow, and follow security best practices to avoid accidental key exposure. It acts with guardrails, confirming before modifying any live resources. This kind of guided, context-aware assistance is the practical benefit. For a site owner, this could mean securely setting up a new cloud-hosted backend service without fumbling through docs alone.
What to Expect Next from the Ecosystem
This first plugin establishes a pattern. The existence of the open Google Agent Skills repository suggests we will see more plugins emerge, covering specific Google Cloud products or advanced infrastructure patterns. The vendor-neutral standard should encourage interoperability. I expect other cloud providers and toolmakers to adopt similar packaging, leading to a more portable agent toolchain where switching between AI assistants doesn’t mean rebuilding your entire tool setup from scratch.
Try It Yourself
You can test this yourself with one of the installation commands. For Antigravity, try:
agy plugin install https://github.com/google/skills/plugins/cloud/google-cloud-developer
After installation, give your agent a simple first prompt like: “What Google Cloud projects do I have access to in my current configuration?” This tests the plugin’s ability to securely interact with your environment and retrieve information.
To explore other available plugins or contribute your own, visit the Google Agent Skills repository (opens in new tab). The source material for this announcement is Google’s official blog post, “Introducing the Google Cloud Developer Plugin for AI Coding Agents (opens in new tab).” Try installing the plugin and giving it a real task from your workflow this week.