Google Goes All-In on MCP Servers — And Exposes the Enterprise Governance Gap

Google's December announcement of managed MCP servers for Maps, BigQuery, Compute Engine, and Kubernetes Engine is a watershed moment for enterprise AI agent infrastructure. For the first time, a hyperscaler is providing production-grade, fully managed MCP endpoints that agents can connect to with a single URL.

The enterprise play is even more significant than the consumer one. Google's Apigee API management platform can translate any standard API into an MCP server, meaning the same API guardrails companies use for human-built applications — key management, quotas, monitoring — can now apply to AI agent tool calls. Google Cloud IAM controls what agents can do with each server. Model Armor provides firewall-style protection against prompt injection and data exfiltration.

This is exactly the right approach to MCP security — and it highlights a critical gap for the rest of the enterprise.

Google can govern MCP servers within its own ecosystem because it controls the infrastructure from agent to tool. But most enterprises do not operate exclusively within Google Cloud. They run a heterogeneous environment with MCP servers connecting to internal databases, third-party SaaS tools, on-premise systems, and multiple cloud providers simultaneously. The governance challenge is not within any single ecosystem; it is across all of them.

When an AI agent in your environment queries a Google BigQuery MCP server, then writes the results to a Slack MCP server, then triggers an action through a custom internal MCP server, the governance question spans three different trust domains. Google's IAM governs the BigQuery call. Slack's permissions govern the message. But nothing governs the end-to-end flow, the data classification decision, or the audit trail that connects them.

This is the enterprise governance gap that Google's announcement makes visible. Individual MCP server providers can secure their own endpoints. But enterprises need a control plane that sits in front of all their MCP servers, regardless of provider, and enforces consistent access control, data policies, and audit logging across the entire tool ecosystem.

Google has built the enterprise model for managed MCP servers. Enterprises need the equivalent model for MCP governance.