Database Investigator is in Preview
Database Investigator is in Preview for all customers monitoring Postgres with Database Monitoring. SQL Server support is in early access: open a SQL Server host within Database Monitoring and click Get early access to request it. Support for other databases covered by Database Monitoring is in development.
Overview
Database Investigator is an AI agent inside Database Monitoring that helps you investigate, optimize, and understand the databases you monitor. It analyzes the telemetry Database Monitoring already collects including health signals, query metrics, explain plans, instance and infrastructure metrics, calling APM services, related incidents, and recent events. For investigations, it returns a structured root cause analysis with concrete remediation steps.
Both DBAs and the platform or application teams that own database-backed services can make use of Database Investigator. Describe an issue in plain language and it runs the investigation. Deep database expertise is not required.
Use Database Investigator to answer questions like:
Why is this database slow?What caused the latency spike on this query in the last hour?Is this regression isolated to one instance or fleet-wide?Which upstream service is driving the workload change?How can I reduce lock contention on this table?
Initiate Database Investigation
Open Database Investigator from any of the following surfaces in Database Monitoring:
- The Investigate button on a database Overview page
- A metric in the Metrics tab of a database host
- A Blocking Queries or Deadlocks panel
- A normalized query in Query Metrics
- A query sample in Query Samples
Database Investigator opens with a suggested starting prompt based on your entry point. You can run that prompt or type your own.
Database Investigator streams its response as it works through the investigation: it states a plan, calls Datadog services to gather evidence, and shares interim findings. The investigation ends with a structured report covering what happened, why, the supporting evidence, the root cause, and recommended fixes.
You can ask follow-up questions at any point to refine the analysis, or paste in external context, such as an execution plan captured outside of Datadog. Database Investigator factors the new information into its next steps.
If a recommended fix involves a SQL change (a new index, a query rewrite), Database Investigator includes the SQL in its response. Database Investigator never connects to your database or runs SQL against it, as recommended changes are for you to review and apply.
Share, save, or escalate findings
After an investigation completes, you can:
- Share the conversation: copy the URL to deep-link any teammate with Database Monitoring access into the same conversation.
- Export to a notebook: capture the investigation in a Datadog notebook for write-up, follow-up, or sharing in a postmortem.
- Create an incident: escalate the findings into a new Datadog incident.
Conversation history
Past conversations are listed in the history view inside Database Investigator. Open a previous conversation to revisit its findings or continue investigating where you left off.
Permissions
To use Database Investigator, your role must have the Database Monitoring Read permission. Database Investigator uses your Datadog role to fetch data, so it can only access the resources you have permission to view.
FAQ
What data does Database Investigator access?
Database Investigator reads the telemetry your account already collects, including query metrics, query samples, explain plans, instance and host metrics. It also reads related APM service dependencies, incidents, events, and metrics in your Datadog organization. It accesses only the resources permitted by your role.
Is data sent to Database Investigator used to train AI models?
Database Investigator operates under zero-retention and zero-training agreements with the third-party AI service providers that power it. Data processed during an investigation is not retained by those providers and is not used to train or improve their models.
Does Database Investigator modify the database?
No. Database Investigator reads observability data and produces recommendations. It does not connect to your database, run SQL against it, or change configuration. SQL changes in its responses are suggestions for you to review and apply.
Which databases are supported?
Postgres is supported in Preview for all Database Monitoring customers. SQL Server is supported in early access. Request access from inside Database Monitoring on a SQL Server host. Support for other databases covered by Database Monitoring is in development.
Are there usage limits?
Database Investigator is included with Database Monitoring. Each Datadog organization can run up to 100 investigations per month.
Does Database Investigator answer general database questions?
Yes. Database Investigator covers diagnosing performance regressions, recommending indexes and optimizations, explaining execution plans, and answering questions about database features, configuration, and best practices. When a question is grounded in your environment, it pulls from the telemetry your account already collects.
Is programmatic access available?
The data sources that Database Investigator uses are also exposed through the Datadog MCP server. You can call them from your own AI tooling.
Further reading
Additional helpful documentation, links, and articles:
