Google unveils BigQuery tools for AI agent data access

Google Unveils New Toolset to Bridge AI Agents with BigQuery Data

Mountain View, CA – July 30, 2025 – Google has launched a new toolset designed to empower enterprises by seamlessly connecting their AI agents with data stored in BigQuery, its cloud-based data warehouse. This development addresses the increasing demand for "agentic applications" – AI systems capable of performing tasks autonomously without human intervention – and aims to provide these agents with the rich context needed for more accurate and relevant responses.

The rise of AI agents is a significant trend in 2025, with many enterprises actively piloting and deploying them to automate repetitive tasks, enhance customer experiences, and augment human capabilities. However, a critical challenge has been enabling these agents to securely and intelligently interact with vast amounts of enterprise data. Google's new toolset directly tackles this by offering a secure and reliable bridge between AI agents and BigQuery.

The newly introduced toolset includes a suite of functionalities that allow AI agents to execute queries within BigQuery and retrieve crucial metadata. Key tools in this set are list_dataset_ids (to get all dataset IDs in a Google Cloud project), get_dataset_info (for detailed dataset metadata), list_table_ids (to list table IDs within a dataset), get_table_info (to fetch metadata for individual tables), and execute_sql (to run SQL queries and retrieve results directly from BigQuery).

This toolset is not a standalone solution; it integrates with Google's existing open-source offerings: the Agent Development Kit (ADK) and the MCP Toolbox for Databases (formerly known as Generative AI Toolbox for Databases). Enterprises can assign the toolset to an agent created within the ADK framework by importing it from the agents.tools module in a Python environment, using the ADK CLI and SDK. The tool_filter parameter also allows for selective exposure of tools to the agent.

Alternatively, the MCP Toolbox for Databases natively supports BigQuery's pre-built toolset. To leverage these tools, enterprises need a Python-supported environment to create an mcp-toolbox folder and install the MCP Toolbox. The MCP Toolbox acts as an open-source server that centralizes the hosting and management of toolsets, enabling agents to act as MCP clients and request tools from the Toolbox, which then handles the complexities of secure connections, authentication, and query execution. Furthermore, the MCP Toolbox deployment mode allows for defining custom SQL tools.

Industry experts, such as Forrester Vice President and Principal Analyst Charlie Dai, believe this integration will significantly accelerate the development of agentic applications. Dai notes that the "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."

Google's move comes amidst a broader industry trend where major data platform providers are focusing on connecting AI agents with enterprise data. Rivals like Databricks, Snowflake, and Teradata have also introduced their own Model Context Protocol (MCP) Servers and related offerings to facilitate AI agent interaction with data stored in their data lakehouses and databases. The Model Context Protocol (MCP) itself, initially rolled out by Anthropic, is emerging as a crucial open protocol for connecting large language models (LLMs) to data sources like BigQuery, enabling them to run SQL queries and interact with projects directly from existing tools.

Google has also been actively enhancing BigQuery with specialized AI agents and an autonomous data foundation. This includes a data engineering agent to assist with building data pipelines and automating metadata generation, a data science agent embedded in Colab notebooks for automating feature engineering and model selection, and a Looker conversational analytics agent that allows natural language queries and provides explanations for its conclusions. These advancements are underpinned by the BigQuery Knowledge Engine, which leverages Google's Gemini models to analyze schema relationships, table descriptions, and query histories, generating metadata and supporting semantic searches. Google has stated plans to expand the newly announced toolset with more functionalities in the future.