Google Unveils MCP Servers: Bridging AI Agents with Services for Enhanced Performance

January 4, 2026

Google lance des serveurs MCP pour connecter les agents IA à ses services

Introduction to Google’s Fully Managed MCP Servers

Google has recently enhanced its cloud and consumer services by integrating the Model Context Protocol (MCP), offering fully managed MCP servers that connect AI agents directly to essential services like Google Maps, BigQuery, Compute Engine, and Kubernetes Engine. This development allows AI agents to harness these services without the need for developers to construct and maintain their own, often fragile, connectors. Over the coming months, Google plans to expand these capabilities across its entire service catalog, incorporating built-in security and governance features through Google Cloud IAM and Model Armor.

MCP: A Standard Protocol for AI Agent Interconnection

The Model Context Protocol, or MCP, is an open-source protocol developed by Anthropic, the creators of Claude, several months ago. It enables artificial intelligence systems to access external data and tools through a unified standard. Without this intermediary layer, developers are tasked with creating specific connectors for each service, a time-consuming and challenging process to sustain. Anthropic has recently handed over the protocol to the Linux Foundation to establish it as an open standard.

Google describes Anthropic’s MCP as akin to a “USB-C for AI,” quickly becoming a standard for linking AI models with data and tools.

MCP Services from Google: Four Services Available at Launch

Google has initially rolled out MCP servers for four of its services. Google Maps provides AI agents with access to up-to-date geospatial data, such as location details, weather forecasts, and routes. This integration enables an assistant to accurately answer queries about distances or recommend eateries without generating fabricated information, according to Google.

BigQuery allows agents to query corporate data directly, interpreting schemas and executing queries while keeping the data within its original environment, thereby reducing security risks associated with transferring sensitive data.

Compute Engine and Kubernetes Engine grant agents the ability to manage the technical infrastructure directly. For instance, an agent can create and resize virtual servers or adjust resources automatically based on workload, all without human intervention.

Secured and Governed Infrastructure

Google’s MCP servers are fortified with multiple layers of security. Google Cloud IAM handles permissions, audit logging provides traceability, and Google Cloud Model Armor serves as a firewall to protect against specific threats to AI agents, such as prompt injection or data exfiltration.

Google’s API management platform, Apigee, converts standard APIs into MCP servers, applying the same security measures to agents as those used for traditional applications.

“We are making Google ready for agents, right from the design phase,” stated Steren Giannini, Product Management Director at Google Cloud.

Staged and Ongoing Deployment

These servers are currently available in a public preview at no additional cost to enterprise clients already using Google’s services. Steren Giannini has indicated to TechCrunch that a general availability launch is expected by early 2026.

Google intends to extend MCP support to its entire catalog in the upcoming months, including Cloud Run, Cloud Storage, AlloyDB, Cloud SQL, Spanner, Looker, Security Operations, and Cloud Logging. As a founding member of the Agentic AI Foundation, the company also commits to contributing to the protocol’s evolution through the open-source community.

Similar Posts

Rate this post

Leave a Comment

Share to...