CLOUD SERVICE · AWS

AWS DevOps Agent - AWS DevOps Agent is an AI-powered operations service that investigates incidents and supports reliability, performance, and release-validation workflows across AWS, multicloud, and on-premises environments.

AWS DevOps Agent correlates resource relationships, telemetry, code changes, and deployment information to help engineers diagnose operational problems and identify improvements. It supports on-demand and scheduled SRE tasks, collaborative investigations, resilience assessment workflows, and a release-management capability available in preview.

Cloud Services Hub →

What is AWS DevOps Agent

Read the extensive description

AWS DevOps Agent helps engineering teams investigate application incidents and identify improvements to reliability and performance. It combines application resource relationships with information from observability tools, runbooks, source repositories, and CI/CD pipelines. This context helps connect operational symptoms with potential causes, including recent deployments and changes in dependent systems. Its scope includes applications running on AWS, in supported external cloud environments, and on premises.

Start an investigation with a specific symptom, an affected application, and a useful time window. For example, when checkout latency increases after a deployment, ask the agent to examine relevant telemetry, resource dependencies, and deployment changes. Review the evidence supporting its conclusions before choosing a mitigation. A correlation between a deployment and an incident is a useful lead, but engineers should still validate the proposed explanation.

Investigation quality depends on the operational context available through configured connections. Connect the relevant observability systems, repositories, delivery tools, and runbooks, and define the application boundary clearly. Missing logs, incomplete dependency information, or inaccessible deployment records can limit the investigation. Scope access to the resources and environments required for the task, and distinguish permissions for investigation from permissions that allow production changes.

Custom SRE agents can run on a schedule within Agent Spaces. A team can create a recurring database health review that examines slow queries and configuration parameters requiring attention. Custom skills and bring-your-own sub-agents provide extension points for specialized workflows. Headless access through MCP and A2A protocols allows compatible assistants and other agents to use DevOps Agent capabilities from existing engineering tools.

Bidirectional Slack communication lets engineers initiate and steer investigations by mentioning the agent in connected conversations. This keeps investigation requests and follow-up questions close to the incident discussion. The AWS Resilience Hub integration provides a separate workflow for reviewing failure-mode assessment findings, implementing mitigations, and examining resilience test results. Teams can use these findings to prioritize preventive work alongside active incident response.

The release-management capability is available in preview and includes release-readiness review and autonomous release testing. Use it as an additional validation layer alongside established tests, deployment approvals, and rollback mechanisms. Before adopting any workflow that can affect production, establish authorization boundaries, evaluate the proposed changes, and define how service health will be checked afterward. Confirm regional availability, connector requirements, and preview restrictions for the capabilities included in the architecture.

Key AWS DevOps Agent Features

Context-aware incident investigation

Correlates telemetry, resource relationships, code, and deployment information to investigate operational problems. Provide an affected application, a symptom, and a time window to focus the investigation.

Reliability and performance analysis

Identifies operational improvements beyond the immediate incident. Validate recommendations against workload requirements and measure their effects through existing testing and change-management processes.

Multicloud and hybrid coverage

Supports investigations across AWS, supported external cloud environments, and on-premises systems. Effective coverage requires access to relevant telemetry and application context from each environment.

Scheduled custom SRE agents

Allows teams to create custom agents within Agent Spaces and schedule recurring work. For example, run a daily database health review covering slow queries and configuration parameters.

Agent extensibility and headless access

Supports custom skills, bring-your-own sub-agents, and access through MCP and A2A protocols. Use these capabilities to incorporate specialized procedures or access operational workflows from compatible assistants.

Bidirectional Slack investigations

Engineers can initiate and steer investigations by mentioning the agent in connected Slack conversations. This keeps investigation requests and follow-up discussion in the team's incident collaboration channel.

Resilience assessment and test review

Integrates with AWS Resilience Hub to review failure-mode assessment findings and resilience test results. Engineers can use this context to work through mitigations for identified weaknesses.

Release readiness and testing preview

Reviews code changes for release readiness and runs autonomous release testing through a preview capability. Retain existing approval gates and rollback procedures when evaluating it.

Custom operational charts and reports

Supports custom charts and reports for operational analysis. Use them to communicate investigation results and recurring health assessments, checking that the underlying data covers the intended application scope.

AWS DevOps Agent Use Cases

Investigate a production regression

Correlate increased errors or latency with recent code and deployment changes. Review supporting evidence before deciding whether to roll back a release or apply another mitigation.

Troubleshoot a hybrid application

Investigate an application whose components span AWS and connected external environments. Include telemetry and dependency context from each environment to reduce gaps in the analysis.

Automate recurring health reviews

Schedule a custom SRE agent to produce an operational assessment, such as a daily database health report. Assign ownership for actionable findings through the team's existing remediation process.

Coordinate investigations in Slack

Start an investigation in a connected Slack conversation and steer it through follow-up messages. Keep operational questions and investigation activity alongside the response team's discussion.

Address resilience findings

Review AWS Resilience Hub assessment findings and test results to understand weaknesses and work through mitigations. Reassess or retest after changes to determine whether they achieved the intended improvement.

Add release validation

Use the release-management preview to review readiness and test proposed changes. Treat its results as an additional input to established deployment controls rather than a replacement for them.

Services AWS DevOps Agent integrates with

A common architecture connects AWS DevOps Agent to the systems that describe an application: observability tools, repositories, deployment systems, and runbooks. CloudWatch telemetry can support AWS workload investigations through configured connections, while relevant CloudTrail events and CodePipeline execution records can supply change context through appropriate access paths. Confirm connector support and permissions explicitly; storing information in an AWS service does not automatically make it available to the agent.Align access with application boundaries and account ownership. Use IAM permissions appropriate to each connection, identify the environments in scope, and distinguish investigation access from permissions required for production changes. For hybrid applications, include external telemetry and dependency information so the investigation does not stop at the AWS boundary.AWS Resilience Hub provides a direct workflow for reviewing resilience findings and test results with the agent. Slack supports collaborative investigations, while MCP and A2A enable access from compatible assistants and other agents. Keep deployment orchestration, approval gates, and rollback mechanisms in the existing delivery architecture, adding preview release validation only after evaluating its requirements and limitations.

AWS CloudTrail

CloudTrail records can provide change evidence when exposed through an appropriate data-access path or supplied as investigation context. For example, compare a permissions change with the start of application access failures. This is a change-correlation pattern, not an assumption that the agent automatically ingests every CloudTrail event.

Open AWS CloudTrail →

Amazon CloudWatch

CloudWatch telemetry can provide operational evidence when made accessible through configured observability connections. For example, an engineer investigating API latency can examine relevant logs alongside error-rate and latency metrics. Confirm supported data types and access scope rather than assuming every log group or dashboard is automatically available.

Open Amazon CloudWatch →

AWS CodePipeline

CodePipeline can remain the deployment orchestrator while relevant execution records and deployed revisions provide investigation context. For example, supply the pipeline execution and revision associated with a release that preceded an error-rate increase. Evaluate the release-management preview separately before incorporating its readiness review or testing into a production delivery workflow.

Open AWS CodePipeline →

AWS Resilience Hub

AWS Resilience Hub integrates with DevOps Agent for reviewing failure-mode assessment findings, implementing mitigations, and examining resilience test results. For example, a platform engineer can launch the agent to review a dependency-related finding and work through a mitigation. Validate the resulting change through the team's resilience assessment and testing workflow.

Open AWS Resilience Hub →