Google is introducing fully managed, remote MCP servers. These will make it easier for developers to enable AI agents to work with Google Maps, BigQuery, Compute Engine, and Kubernetes.
Google’s existing API infrastructure now supports MCP and provides a uniform layer across all Google and Google Cloud services. Developers can point their AI agents or standard MCP clients to a globally consistent and enterprise-ready endpoint for Google services.
An essential addition is the expansion to the broader enterprise stack via Apigee. This enables companies to unlock and manage their own developer-built APIs as well as external third-party APIs. The incremental rollout of MCP support starts with four services.
Google Maps Grounding Lite connects AI agents to trusted geospatial data and provides access to up-to-date information about locations, weather forecasts, and route details. An AI assistant can use this to answer questions such as “How far is the nearest park from this rental property?” or “What child-friendly restaurants are there near our hotel?”
BigQuery, Compute Engine, and Kubernetes
The BigQuery MCP server enables agents to interpret native schemas and execute queries on enterprise data without the security risks or latency of moving data to context windows. Data stays in place and remains subject to governance.
Google Compute Engine (GCE) enables agents to manage infrastructure workflows autonomously. By offering capabilities like provisioning and resizing as discoverable tools, agents can handle everything from initial builds to day-2 operations, dynamically adjusting to workload demands.
The Google Kubernetes Engine (GKE) MCP server provides a structured, discoverable interface that enables agents to communicate reliably with the GKE and Kubernetes APIs. This eliminates the need to parse text output or string together complex CLI commands.
Security and observability are built in via Cloud API Registry, Apigee API Hub, Google Cloud IAM, and audit logging. Google Cloud Model Armor is designed to protect against advanced agentic threats, such as indirect prompt injection.
Tip: Google lets AI agents do the data work in BigQuery and Looker