Google’s new toolset to help connect AI agents to BigQuery

Google has come out with a new toolset that will allow enterprises to connect their AI agents to data stored inside BigQuery in the wake of rising demand for agentic applications.

Agentic applications, which can perform tasks without manual intervention, have caught the fancy of enterprises as they allow them to do more with constrained resources.

However, enterprises have been looking at ways to provide additional context to the AI-based agents inside these applications to help them generate more accurate responses to a user request, and Google claims that the toolset will serve that purpose.

The new toolset contains tools that will allow AI agents to execute queries inside BigQuery as well as fetch metadata from the cloud-based data warehouse.These tools include the list_dataset_ids tool, which retrieves all dataset IDs within a Google Cloud project; get_dataset_info tool, which provides detailed metadata about a specific dataset; list_table_ids tool, which lists all table IDs within a dataset; get_table_info tool, which fetches metadata for individual tables; and execute_sql tool, which allows users to run SQL queries directly in BigQuery and retrieve results.

The toolset, nonetheless, cannot be implemented in itself, and enterprises need to use the toolset in combination with Google’s open source offerings — Agent Development Kit (ADK) and its MCP Toolbox for Databases, formerly called Generative AI Toolbox for Databases, to be able to connect their agents to BigQuery.

If an enterprise wants to use the ADK, it needs to assign the toolset to an agent created within the framework itself.

The toolset can be assigned by importing it from the agents.tools module within an environment running Python via the ADK command line interface (CLI) and SDK, Google explained in a blog post.

Enterprises will also have the option of using the tool_filter parameter to filter the tools they want to expose to the agent, it added.

On the other hand, the MCP Toolbox for Databases natively supports BigQuery’s pre-built toolset, and to access these tools, enterprises need a Python-supported environment to create a new mcp-toolbox folder in the same directory as their ADK-developed agentic application, and then install the MCP Toolbox.

Google also provides an option in the MCP Toolbox deployment mode for enterprises to define their respective custom tools in SQL.

Forrester vice president and principal analyst Charlie Dai feels that the toolset’s integration will go a long way in accelerating the development of agentic applications.

“Google’s ADK and MCP integration provides pre-built frameworks to connect AI agents directly to BigQuery data. This eliminates custom integration work, reducing development overhead, and enables agents to leverage enterprise context for accurate responses,” Dai said.

Google isn’t alone in the race to connect AI agents with enterprise data.

In the last few months, BigQuery rivals, such as Databricks, Snowflake, and Teradata, have all introduced MCP Servers and other MCP-related offerings to help enterprises connect AI agents to data stored inside their data lakehouses and databases. Google has said that it is planning to add more tools to the newly announced toolset, but didn’t provide a timeline for their release.