Google launches AI agents for database setup & ops
Wed, 5th Aug 2026 (Yesterday)
Google has introduced two AI database agents for setup and operations management as part of its broader Agentic Data Cloud push.
The new products are the Database Onboarding Agent, which handles initial database setup and configuration, and the Database Observability Agent, which focuses on monitoring, troubleshooting and maintenance after deployment.
The observability tool is designed for Site reliability engineers, DevOps teams, database administrators and developers who need to investigate issues through natural language prompts rather than manual checks across multiple systems.
It combines telemetry from Database Insights, Cloud Monitoring, Cloud Logging and Cloud Trace to identify the causes of issues such as latency spikes, lock contention and query hotspots. In some cases, it can also recommend corrective actions and carry out approved changes.
One example Google gave was a recommendation to enable connection pooling for a Cloud SQL instance when the agent detects a bottleneck. The system presents the rationale and expected effect before the user decides whether to proceed.
Workflow tools
The observability functions are available across several interfaces, including Cloud Assist chat, the Google Cloud console, command-line tools, managed context protocol servers and third-party development tools. This allows developers and operations teams to use the same system within existing workflows rather than switch to a separate product.
The agent is also integrated with Database Centre, allowing fleet-wide queries across an organisation's database estate. A user could ask which databases consumed the most CPU over a set period and receive a summarised response.
Some observability features remain in preview with select customers, including in-product investigations and validated remediations. These functions are intended to move the tool beyond diagnosis into supervised action.
The onboarding agent addresses an earlier stage of the database lifecycle. Instead of reviewing product documentation and mapping requirements manually, users can describe workload needs in natural language and receive a recommendation on which managed database service to use.
The onboarding tool considers factors such as performance needs, scale, data type and reliability requirements. It is also intended to understand technical measures including IOPS, latency thresholds and replication lag when making recommendations.
Database choices
After a user selects a service, the onboarding agent can generate the commands needed to provision and configure a database instance. This is intended to reduce the scripting and manual setup work typically involved in early deployment.
The announcement reflects a broader shift among cloud providers to apply generative AI to infrastructure administration, aiming to lower the skill barrier for routine operations while shortening the time needed to diagnose faults.
Database management has traditionally required specialist knowledge across architecture planning, schema design, indexing, query tuning and monitoring. As applications scale, identifying the exact source of poor performance often involves reviewing metrics, logs and traces across multiple products.
Google is positioning the new agents as a way to combine those sources and reduce repeated manual investigation. The observability tool can produce root-cause analysis in minutes by linking telemetry across services.
The observability agent supports several of Google's managed database products. These include Cloud SQL, where it can examine load, query performance and connectivity issues; Spanner, where common tasks include investigating read and write latency; AlloyDB, where it can look into replica lag as well as load and query performance; and Bigtable, where it focuses on read and write latency.
The broader set of services covered by the two agents includes AlloyDB, Bigtable, Cloud SQL for PostgreSQL, MySQL and SQL Server, Firestore, Memorystore and Spanner. Access is being provided through Gemini Cloud Assist.
Observability functions are also being made available as managed context protocol tools through the Database Insights MCP Server and Database Centre MCP Server, extending access to developers using external environments and IDEs.
Google says the onboarding and observability agents are intended to place database guidance inside chat, console and developer interfaces, with the observability agent able to suggest and, with approval, execute validated remedial actions.