Google Cloud launches remote MCP server for Vertex AI Agent Platform

Full analysis

Google Cloud has released the Gemini Enterprise Agent Platform remote MCP server, a bridge that lets external AI agents (built in tools like Anthropic's Claude Code or other IDEs) securely connect to resources inside a Google Cloud environment. The server uses the open MCP (Model Context Protocol) specification — a standardized way for AI agents to call tools and access data — so agents can invoke models from Model Garden, pull shared prompt templates, or manage Notebooks without leaving their development environment.

The server activates automatically when developers enable the Gemini Enterprise Agent Platform API and enforces access controls via native Cloud IAM Deny policies, addressing the tension between developer speed and IT governance. A centralized Agent Registry lets teams catalog and govern their inventory of tools, skills, and AI capabilities. This follows Google's earlier announcement that over 50 Google-managed MCP servers are available, signaling continued investment in MCP as the connective tissue for enterprise agent infrastructure.