Google Cloud expanded support for agentic AI standard model context protocol (MCP) to all Google and Google Cloud services currently using the standard.

MCP was launched in November 2024 by Anthropic as an open standard for the AI development community. It's used as a plug-and-play bridge for large language models (LLMs) and AI agents to connect with external tools.

Google has launched managed remote MCP servers, eschewing the need for developers to manually build, install, and host their own local MCP solutions. Developers will also be able to direct their custom agents or standard MCP clients such as Gemini CLI to a globally uniform and enterprise-ready endpoint for Google and its Cloud services.

Google is also providing official remote MCP support for the BigQuery MCP, Google Compute Engine (GCE), and Google Kubernetes Engine (GKE).

The BigQuery MCP integration enables agents to natively interpret schemas, execute queries, and forecast data without moving data into context windows and without any compromise of data residency requirements.

The GCE update allows agents to perform infrastructure operations, such as provisioning and resizing resources.

“The GKE MCP server exposes a structured, discoverable interface that allows agents to interact reliably with both GKE and Kubernetes APIs, eliminating the need to parse brittle text output or string-together complex CLI commands,” added Michael Bachman, VP/GM, Google Cloud.

In regards to safety concerns with MCP use – such as AI agents having the autonomy to possibly take incorrect network actions – Google stressed its managed MCP servers integrated with Google Cloud's identity and access management tools (IAM) for access control, while extra security guardians prevent threats such as indirect prompt injection, a danger intrinsic to the looming threat of so-called shadow AI.

The tech giant is also launching a new API registry and hub for MCP tool discovery, as well as MCP support for its API management platform Apigee.

These directories run in parallel to efforts such as the Agntcy project, a collaboration between Cisco, Google Cloud, Oracle, Red Hat, and Dell Technologies, which was donated to the Linux Foundation (LF) as well as the Agentic AI Foundation (AAIF), LF’s neutral AI development infrastructure, which comes anchored by standards including MCP.