Google Launches Managed MCP Servers for AI Agents

Addressing the Challenges of AI Agent Integration

Artificial intelligence agents are increasingly marketed as solutions for diverse tasks, ranging from travel planning to complex business problem-solving. However, effectively connecting these agents with external tools and data sources beyond their native chat interfaces has proven to be a significant hurdle.

Traditionally, developers have relied on creating and maintaining custom connectors, a process that is often fragile, difficult to scale, and introduces complexities in governance and oversight.

Google's Solution: Managed MCP Servers

Google proposes a solution through the launch of fully managed, remote Model Context Protocol (MCP) servers. These servers aim to simplify the integration of Google and Cloud services – including Maps and BigQuery – with AI agents.

This initiative builds upon the recent release of Google’s Gemini 3 model, focusing on combining enhanced reasoning capabilities with more reliable connections to real-world tools and data.

Streamlined Developer Experience

According to Steren Giannini, Product Management Director at Google Cloud, the goal is to make Google services inherently “agent-ready.”

Instead of dedicating weeks to connector setup, developers will be able to integrate services by simply providing a URL to a managed endpoint.

Initial Server Availability and Use Cases

The initial launch includes MCP servers for Maps, BigQuery, Compute Engine, and Kubernetes Engine.

Practical applications include an analytics assistant directly querying BigQuery, or an operations agent interacting with infrastructure services.

For Maps, utilizing the MCP server ensures agents have access to current, accurate location data for trip planning and similar applications, surpassing the limitations of the model’s pre-existing knowledge.

Public Preview and Cost

While Google intends to extend MCP server availability across its entire suite of tools, they are currently being released under a public preview. This means they are not yet fully covered by standard Google Cloud terms of service.

However, the servers are offered at no additional cost to enterprise customers who already subscribe to Google services.

Google anticipates a general availability release in the coming year, with a continuous rollout of new MCP servers on a weekly basis.

Understanding the Model Context Protocol (MCP)

The Model Context Protocol (MCP), originally developed by Anthropic approximately a year ago, is an open-source standard designed to facilitate connections between AI systems and data/tools.

This protocol has gained widespread adoption within the agent tooling ecosystem, and Anthropic recently contributed MCP to a new Linux Foundation fund dedicated to open-sourcing and standardizing AI agent infrastructure.

“The strength of MCP lies in its standardization,” Giannini explained. “A server provided by Google can connect to any compatible client.”

MCP Clients and Compatibility

MCP clients are the AI applications that interact with MCP servers to access the tools they offer.

Google’s own Gemini CLI and AI Studio serve as MCP clients, but the protocol has also been successfully tested with Anthropic’s Claude and OpenAI’s ChatGPT.

The Role of Apigee in Enterprise Integration

Google emphasizes that this initiative extends beyond simply connecting agents to its services.

Apigee, Google’s API management product, plays a crucial role in enterprise integration by allowing companies to leverage existing API keys, quotas, and traffic monitoring capabilities.

Apigee can effectively “translate” standard APIs into MCP servers, enabling agents to discover and utilize endpoints like product catalog APIs, all while maintaining existing security and governance controls.

Security and Governance Features

Google’s new MCP servers are secured by Google Cloud IAM, which precisely defines an agent’s permissible actions.

Furthermore, Google Cloud Model Armor acts as a firewall specifically designed for agentic workloads, protecting against threats such as prompt injection and data exfiltration.

Administrators also benefit from comprehensive audit logging for enhanced observability.

Future Expansion and Developer Focus

Google plans to broaden MCP support to encompass additional services in areas such as storage, databases, logging, monitoring, and security in the coming months.

“Our aim is to provide the foundational infrastructure, relieving developers of that burden,” Giannini concluded.