Google is now supporting the Model Context Protocol (MCP) in its cloud infrastructure. Fully managed MCP servers are available for Google Maps, BigQuery, Compute Engine, and Kubernetes Engine. The servers are currently in a phased rollout (Public Preview).

MCP is an open protocol developed by Anthropic that acts as a standardized interface between AI agents and data sources. It is often referred to as "USB-C for AI" because it creates a universal connection layer. Previously, developers had to program custom connections for each service, which was error-prone and time-consuming.

Managed Servers Instead of Local Installation

Users of Google Cloud services no longer need to install and operate their local MCP servers. Instead, they can use globally available, Google-managed endpoints.

The MCP servers are integrated into Google’s existing security and governance infrastructure. Access is controlled via Google Cloud IAM, and all actions are logged in the audit log. Additionally, Google Cloud Model Armor is intended to protect against agentic threats such as indirect prompt injections or attempted data exfiltration. Administrators can define policies at the organization level and selectively enable or disable individual services.

Expansion to More Services Planned

Google plans to significantly expand MCP support in the coming months. Cloud Storage, various database services, and logging, monitoring, and security services are on the roadmap. In the long term, Google intends to integrate MCP into many more of its cloud services, according to the announcement.

Via Apigee, Google is also extending MCP integration to proprietary APIs and third-party APIs. With Cloud API Registry and Apigee API Hub, developers can publish and manage their APIs as MCP tools. This allows AI agents to access internal systems and data sources as well. Furthermore, Anthropic, the original developer of MCP, is working closely with Google within the new Agentic AI Foundation on the further development of the specification.

MCP is already used by popular AI services such as ChatGPT, Claude, Microsoft Copilot, and Visual Studio Code (with GitHub Copilot). Major cloud providers such as AWS, Microsoft, and Cloudflare are main members of the Agentic AI Foundation, which is further developing MCP, making it the de facto standard for agentic AI applications.

(fo)

Don’t miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.