Try it free

Set up SRE Agent

  • Latest Dynatrace
  • How-to guide
  • 3-min read
  • Published Aug 05, 2026
  • Preview
Table of contents (Preview)
  • Agentic workflows
  • Set up Alert Reduction Agent
  • Set up Database Operations Agent
  • Set up Infrastructure Optimization Agent
  • Set up Kubernetes Troubleshooting Agent
  • Set up Mobile Crash Remediation Agent
  • Set up Threat Triage Agent
  • Set up Vulnerability Verification Agent
  • Set up Security Insights Report Agent
  • Set up Security Association Agent
  • Set up Kubernetes Operations Agent
  • Set up SRE Agent

The Site Reliability Engineering (SRE) Agent is a Dynatrace agentic workflow that uses agentic AI to analyze triggered problems and surface root cause, real-user impact, and remediation suggestions directly as annotations. These annotations are visible and can be accessed on the problem details page.

When a problem matches the configured trigger filter, the workflow immediately posts a progress annotation, and then invokes the SRE Agent. The agent correlates metrics, logs, events, deployment events, and traces scoped to the problem. Based on the correlated data, the agent provides a structured analysis that contains root cause, key findings, conclusion, and recommendations, added to the problem as a follow-up annotation.

Prerequisites

  • Dynatrace Intelligence (Preview) app
  • Workflows Workflows

Get started

1. Configure the problem trigger

  1. Open the SRE Agent workflow in Workflows Workflows.
  2. Select the Problem trigger step to review the default problem filter configuration:
    • Event state: active. This means that the workflow only fires for an open problem.
    • Event category: Error, Availability.
    • Severity: Any. This means that the trigger creates an alert for any severity level event.
    • Minimal duration: None. This means that there's no minimal time that a problem has to stay open before the trigger fires.
    • Additional custom filter: dt.davis.is_duplicate == false and not isNull(root_cause_entity_id).
    • Initial root cause analysis: enabled by default. This means that the workflow runs only after the Dynatrace agentic AI has finished its initial analysis.
  3. In the Affected entities field, select whether you want to include all entities or filter them by tags. For example, include all entities.
  4. Adjust the filter to suit your needs. To limit invocations to high-value incidents, we recommend doing the following:
    • In the Event category, select Slowdown in addition to Error and Availability.
    • Set Severity field to only trigger for critical events. For example, Critical (1) or Major (2).

We recommend starting with narrow trigger conditions and expanding the filter after you validate the agent's output quality.

2. Review the workflow tasks

The workflow contains three actions that run in sequence:

  1. progress_msg: posts an initial annotation on the problem details page as soon as the workflow starts, letting your operators know that analysis is in progress.

  2. call_sre_agent: calls the Prompt agentic AI action with access to two MCP tools:

    • execute-dql: runs DQL queries to retrieve logs, events, metrics, and traces related to the problem.
    • get-problem-by-id: retrieves full problem details, including affected entities and root cause entity.

    The agent is constrained to a maximum of five tool calls and produces a response of up to 1,000 words, structured as follows:

    • Root-Cause Analysis: an executive summary of the root cause.
    • Key findings: a list of key findings.
    • Conclusion: a summarized conclusion of the analysis.
    • Recommendations: recommended remediating measures.

  3. agent_report: writes the agent's output as an annotation and sends it back to the problem. The annotation source is labeled as SRE Agent.

If you want to limit or expand the data sources available to the agent, you can:

  1. Go to the call_sre_agent action.
  2. On the Input tab, select or deselect additional tools in the Tool selection section.

3. Activate and validate the workflow

  1. Set the workflow trigger to Active and save the workflow.
  2. To validate the setup, create or replay a synthetic problem that matches your trigger filter.
  3. Open the problem details page and confirm that the two following annotations appear:
    • An initial progress annotation from the SRE Agent.
    • A follow-up annotation that contains the structured analysis result.
  4. Review the annotation for accuracy, clarity, and actionable guidance. If the analysis is missing evidence, try widening the lookback window or increasing the trace sample size in the agent prompt.

Best practices

To ensure that your SRE Agent workflow remains optimized and produces quality output, we recommend adopting the following best practices:

  • Track how often the workflow triggers. Since the workflow enforces an hourly execution limit of 1,000 runs, we recommend starting with strict trigger filters to control invocation volume and cost.
  • Periodically review annotation quality and adjust the trigger filter or agent prompt as your environment evolves.
  • Add an invocation limit or a time-of-day condition to the workflow to prevent the agent from being triggered too often.
  • Ensure that the workflow runs under a Dynatrace service account and not a personal user account, and avoid including any PII or other sensitive information in your prompts or annotations.
Related tags
Dynatrace PlatformGenerative AI for Workflows