The rise of AI agents is fundamentally disrupting applications and analytical systems. Generic AI platforms don't usually have access to the context stored within enterprise databases. This is because traditional data architectures often lack context for agents across the data estate, which can lead to agents being inaccurate. They’re also prone to security gaps due to a lack of granular access controls.
Google’s Agentic Data Cloud is an AI-native system of action that includes both operational and analytical systems. By infusing AI across the entire stack — from custom silicon to frontier Gemini models — we provide a deterministic, template-driven developer framework that allows agents to ground their reasoning in real-time enterprise data with near-100% accuracy, as well as unified governance.
Today, we’re making it easier to develop agents, with a whole host of new data agents and tools: for business analysts within Conversational Analytics; for data scientists, engineers, and database admins with a series of Google-built Data Agents that provide greater automation and intelligence; and finally, for developers, with Data Agent tools that help you better integrate with today’s open agentic ecosystem.
To support developers building agents using natural language, we’re announcing expanded support for Conversational Analytics across Data Cloud.
Conversational Analytics in BigQuery, in preview, integrates a sophisticated AI reasoning engine directly into BigQuery Studio, helping data and business teams go beyond writing manual SQL, leveraging business context to ground answers using multimodal synthesis and deep-dive research. Agentic workflows, in preview for select customers, automate root-cause analysis, and schedule actions — turning enterprise data into proactive, actionable intelligence.
Conversational Analytics in Lakehouse, now in preview, extends the Lakehouse unified infrastructure, so users can query distributed data lakes across AWS, Azure, and Google Cloud using natural language. This makes it possible to combine insights across cloud platforms without moving a single byte of data.
**Conversational Analytics in AlloyDB, Spanner, and ** Cloud SQL, now in preview, supports out-of-the-box conversational AI, making data accessible for everyone. AlloyDB, Spanner, and Cloud SQL users can start natural-language conversations with their databases to gain visibility into their real-time operational data and capture analytical insights.
To help data professionals move from reactive data management to proactive intelligence, and business analysts better interact with their dashboards, we’re announcing a new set of data agents that bring automation, intelligence, and natural language capabilities into their daily workflows.
**Data Engineering Agent, **now generally available, automates the heavy lifting of building and maintaining data pipelines. It transforms natural language requirements into optimized SQL or Python code for BigQuery and Dataflow, while proactively identifying and fixing pipeline breaks. By suggesting schema improvements and partitioning strategies, it ensures your data foundation is scalable, reliable, and performance-tuned without manual trial and error.
**Data Science Agent, **now in preview, accelerates the path from raw data to production-ready models. It assists data scientists by suggesting relevant features, generating boilerplate notebook code, and automating the technical documentation process.
Database Observability Agent, in preview with select Cloud SQL, AlloyDB, Spanner, and Bigtable customers, proactively monitors database performance and continuously identifies potential issues before they escalate. It then delivers intelligent recommendations and multi-turn remediation workflows for fast, comprehensive troubleshooting and optimization. It provides performance analytics for the entire database fleet, helping you quickly identify performance optimization opportunities across databases.
Looker Dashboard Agent, now in preview, enables conversational interaction with data within dashboards. Users can ask natural language questions and receive context-aware answers within the dashboard. This feature also provides AI-generated summaries that highlight key takeaways and insights from the dashboard.
Data Insights Agent, now in preview, provides unified insights into your data assets in Gemini Enterprise, by simultaneously querying structured sources like BigQuery and Snowflake alongside unstructured data like meeting notes and public web info. It functions as a quick-response engine for everyday business users, synthesizing information across the Workspace ecosystem (Docs, Sheets, Drive) and third-party apps like Jira and HubSpot. The agent features rich, interactive visualizations and learns continuously to align with user preferences over time.
Deep Research Agent, now in preview, uses the Knowledge Catalog to solve high-stakes, multi-layered business problems. It moves beyond simple search to build comprehensive research plans that synthesize intelligence from internal documents, BigQuery tables, and the public web. The result is a detailed report with dynamic visualizations and verifiable citations, that respect enterprise privacy and user permissions all the while.
Open-source standards for agentic development provide developers building AI applications and custom agents with a unified framework to access data and tools consistently and securely. Today, we are announcing the following tools to help ground your agentic development initiatives:
**Data Agent Kit: **now in preview, provides a standardized suite of skills and tools directly within preferred developer environments (IDE/CLI), empowering data practitioners to discover, transform, and action data at scale using the prescriptive guidance from the Agentic Data Cloud capabilities.
**Managed MCP Servers for Databases, **now generally available for AlloyDB, Spanner, Cloud SQL, Bigtable, and Firestore, fully manages the infrastructure required to connect AI models securely to your data, so you don’t have to host, secure, or scale MCP servers yourself. Now, developers can provide their agents with up-to-date context from across our database portfolio, so that your AI models can reason and act upon your most up-to-date enterprise data.
Managed MCP Server for Looker, now in preview, allows any MCP client or agent platform to query Looker's semantic models, extending governed BI insights across third-party applications.
**MCP Toolbox for Databases 1.0, **now generally available, has achieved a major stability milestone, giving you the confidence to build production applications. We also overhauled the documentation, making the platform significantly more approachable for both human developers and autonomous agents.
QueryData for Cloud SQL, AlloyDB, and Spanner, now in preview, turns natural language questions into database queries. It’s built natively into these databases, and provides near-100% accuracy for natural language to SQL conversions through metadata, query examples, and evals.
Universal Commerce Protocol (UCP) Analytics powered by BigQuery, now in preview, enables merchants and developers to stream real-time events from UCP directly into BigQuery (see sample). This integration provides out-of-the-box observability for agentic commerce, allowing teams to monitor conversion funnels, track automated checkout performance, and identify system errors. By standardizing these metrics within BigQuery, businesses can bridge the gap between AI-driven transactions and existing business intelligence workflows.
Details on how to access the new agents and tools can be found from each of the documentation links on this page. Data agents are also available through Gemini Enterprise and the Google Cloud console.