Google Cloud launches AI coding plugin for developers

Google Cloud has launched a developer plugin for AI coding agents, adding an installable bundle designed to help agents work with Google Cloud tools and documentation.

The plugin, called google-cloud-developer, is being published through the Google Agent Skills repository and follows the Agent Plugins specification, which Google described as an open standard for packaging agent skills and Model Context Protocol servers.

The release reflects a broader effort to simplify how developers connect AI coding assistants to cloud workflows. Instead of managing separate skills and tool connections one by one, developers can use plugins to group related functions into a single package.

The first release is a foundational plugin for general Google Cloud use. It is designed to help agents retrieve Google Cloud-related skills, use official documentation, and carry out programmatic interactions with Google Cloud services.

Tool grouping

One issue Google is trying to address is what it called the coupling problem: coding agents perform better when they can use several related tools together. In practice, that can mean combining product knowledge, recommended workflows, and access to a live environment when analysing infrastructure or setting up services.

By packaging those elements together, the plugin system aims to reduce the overhead of maintaining separate dependencies across different AI assistants. Google said the standard provides a common manifest and directory structure, so developers do not need to build different wrappers and configurations for each tool.

The google-cloud-developer plugin focuses on basic tasks involved in working with Google Cloud, including authentication, authorisation, project management, and guardrails for gcloud command-line operations.

It also includes configuration for the Developer Knowledge MCP server, which is intended to give agents access to current Google developer documentation. This allows an AI coding assistant to draw on official material while using bundled tools tied to Google Cloud administration.

Onboarding use

Google illustrated the plugin with a project onboarding scenario. In that example, an agent helps a user new to the platform set up an account, create a first project with billing, and authenticate a local machine so a script can call APIs through a service identity rather than an individual user account.

According to Google, the agent can check a live environment in the background for prerequisites such as command-line availability and any existing projects or organisations. It can also review identity and access management practices and present a workflow before making changes to resources.

That emphasis on review steps and operational guardrails points to a practical concern surrounding AI coding tools in cloud environments. As developers give agents more direct access to infrastructure, controls around credentials, permissions, and command execution become more important.

Open standard

Google said its plugin complies with the Agent Plugins specification, which it described as vendor-neutral. It argued that interoperability matters because developers increasingly use a range of coding environments and assistants rather than a single tool.

By adopting that specification, Google is also signalling that it wants its cloud tools to be accessible across third-party agent ecosystems. It highlighted support for installation in Antigravity CLI, Claude Code, and Codex CLI, suggesting it is targeting developers who already switch between multiple AI-assisted coding interfaces.

The release also expands the role of the Google Agent Skills repository, which is positioned as a distribution point for plugins beyond the flagship google-cloud-developer package. Developers using the repository should receive consistent access to plugin-based tools across different AI coding environments.

Competition among cloud providers and AI tool vendors has increasingly centred on how easily developers can connect language models to operational systems. In that context, packaging documentation access, administrative workflows, and cloud interaction into a single plugin may reduce friction for teams experimenting with agent-based development.

The launch also highlights how documentation is being reworked for AI consumption. Instead of serving only as a resource for human readers, official technical documentation is being bundled into systems that can ground an agent's responses and actions while it works through setup or operational tasks.

For Google Cloud, the initial focus is on foundational developer tasks rather than highly specialised product functions. That includes project creation, identity configuration, and safe command-line use, which are often among the first hurdles for developers adopting a cloud platform.

The plugin is available through the open Google Agent Skills repository and can be installed in supported command-line environments using the standard plugin layout.