Google is embracing the MCP standard in its database services to connect AI agents directly and securely to the right data sources.
Google Cloud is expanding its support for the Model Context Protocol (MCP) with managed servers for AlloyDB, Spanner, Cloud SQL, Firestore, and Bigtable, among others. Additionally, the company is introducing a new Developer Knowledge MCP server that links development environments to Google documentation.
With this expansion, Google Cloud aims to allow developers to connect AI agents directly and securely to operational data sources, as stated in a blog post. The new MCP servers run in Google Cloud and provide a unified interface for Gemini and other MCP-compatible clients.
Developers do not need to deploy any infrastructure themselves. They simply configure an MCP endpoint in their agent settings.
Access to databases
The expansion includes support for AlloyDB, Spanner, and Cloud SQL. This allows agents to create schemas, analyze slow queries, and perform vector searches, among other tasks. Through Spanner, they can combine relational, graph, and semantic data using SQL and GQL queries.
For NoSQL workloads, Google Cloud is also adding Bigtable and Firestore. Bigtable is often used for time-series data and large-scale integration platforms. Firestore focuses on mobile and web applications. For example, agents can check statuses or request order information using natural language.
According to Google Cloud, the new approach makes it possible to scale agent-based applications without additional operational overhead. Security is handled through identity management: agents only gain access to explicitly permitted tables or views, and all interactions are recorded in logs.
Support for Gemini and other agents
The announcement coincides with the launch of Gemini 3, which is available via Vertex AI and Gemini Enterprise. The model offers extensive reasoning capabilities but requires access to external systems to function as a full-fledged agent.
Google Cloud demonstrates how an agent uses natural language to migrate a local application to a managed PostgreSQL instance in Cloud SQL. In doing so, the agent uses both the Cloud SQL MCP server and the Developer Knowledge MCP server, which provides relevant documentation.
read also
What is MCP and why is it Suddenly so Popular?
Because the servers follow the open MCP standard protocol, they also support third-party agents, such as Anthropic’s Claude. Users can connect to their Google Cloud databases via a custom connector. In the coming months, Google Cloud plans additional MCP support for services such as Looker, Database Migration Service, and Pub/Sub.