AWS DevOps Agent
Amazon Web Services · AWS DevOps Agent: Autonomous Incident Investigation and SRE Agent
AWS DevOps Agent is an AWS operations agent that investigates incidents around the clock, finds likely root causes across your observability, code, and ticketing tools, and recommends mitigations. It suits SRE and platform teams running production workloads on AWS, other clouds, or on-premises.
Best for
SRE, platform, and DevOps teams running production workloads on AWS, and often other clouds or on-premises, who already use observability and incident tools and want faster, always-on incident investigation
Not ideal for
Small teams with little telemetry or incident volume, organizations that need a fixed monthly price, and teams that need a self-hosted operations agent
Who it's for
SRE, platform, and DevOps teams operating production systems on AWS, multicloud, or on-premises
Capabilities
- Autonomous incident investigations that start from alerts or on request, with an investigation journal and root-cause analysis
- Mitigation recommendations for each investigation, plus opt-in directed actions such as running a mitigation plan (actions that create or modify resources are disabled by default)
- Custom agents that run on a schedule or from webhook events
- Investigates incidents in Azure workloads and, through MCP, on-premises applications
- Triage agent that assesses severity and detects duplicate incidents
- Proactive prevention that analyzes past incident patterns and recommends fixes
- On-demand SRE tasks through chat, covering AWS resources, metrics, alarms, deployment history, and incident patterns
- Bidirectional Slack investigations: @mention the agent in a private channel and keep findings in one thread
- Observability integrations with Amazon CloudWatch, Datadog, Dynatrace, Grafana, New Relic, and Splunk
- Code and CI/CD integrations with GitHub, GitLab, and Azure DevOps, plus code indexing across repositories
- Incident management integrations with PagerDuty, ServiceNow, and Slack, plus Amazon EventBridge
- Learned and custom skills for organization-specific knowledge, and custom tools through A2A and MCP servers
- AWS CLI, SDKs, and CloudFormation, CDK, and Terraform support for skills, custom agents, and triggers
- Release Management (preview): release-readiness reviews, cross-repository dependency mapping, and change-specific test plans
Limitations
- Billed per agent-second ($0.0083, about $30 per agent-hour), so long or frequent investigations add up and are harder to predict than a flat fee
- Queries against connected AWS services, such as CloudWatch Logs Insights, are billed separately at their standard rates
- Agent Spaces can be created in 11 AWS Regions (up from six at GA), and an Agent Space's data stays in the Region where it is created. The agent can still investigate resources in any Region
- Release Management is in preview, free during the preview, and available only in US East (N. Virginia)
- Results depend on the observability, code, and ticketing data you connect. Teams with little telemetry will get less from it
- Proprietary AWS service. No self-hosted or open-source version is documented
Use cases
- Letting the agent start an investigation automatically when a CloudWatch or Datadog alarm fires overnight
- Running an incident from a Slack channel by @mentioning the agent and asking follow-up questions in the thread
- Asking which recent deployment most likely caused an error-rate spike
- Reviewing the agent's prevention recommendations after a run of similar incidents
- Checking a release for readiness and cross-repository breaking changes before deploying (preview)
Our take
AWS DevOps Agent is an incident investigator rather than a general coding tool. Its value comes from reading across the tools a team already uses, including many that are not AWS services, and starting an investigation as soon as an alert fires, so findings are ready when the on-call engineer joins. The per-second pricing is fair for occasional incidents but needs watching on noisy systems, and AWS Support credits matter a lot to the real cost. Teams should start with the free trial on one Agent Space and compare its root-cause calls against a few past incidents before relying on it.
Who should use it
On-call SRE and platform teams with real incident volume and existing observability tooling, especially AWS customers on Business Support+, Enterprise Support, or Unified Operations plans whose credits offset usage.
Who should skip it
Small teams with few incidents or little telemetry, organizations whose data-residency rules require a Region outside the 11 supported ones, and teams that need a self-hosted operations agent or a fixed monthly price.
Strengths
- Starts investigations automatically and works around the clock
- Wide set of built-in observability, code, and incident integrations beyond AWS's own tools
- Slack-native investigations keep the whole team in one thread
- Extensible through custom skills, A2A, and MCP servers
- Two-month free trial and AWS Support plan credits lower the cost of trying it
Weaknesses
- Per-second pricing is hard to predict during busy incident periods
- Connected AWS service queries are billed on top
- Release Management is still in preview and only in US East (N. Virginia)
AWS DevOps Agent pricing
Free trial
Free for 2 months
- For new AWS DevOps Agent customers, starting with their first operational task
- Up to 10 agent spaces
- 20 hours of investigations, 15 hours of evaluations, and 20 hours of chat tasks per month
Pay-as-you-go
$0.0083 per agent-second
- Same rate for investigations, evaluations, and on-demand SRE tasks
- Billed only for time the agent spends working
Note: AWS Support plans include monthly DevOps Agent credits based on the prior month's support charges: 100% for Unified Operations, 75% for Enterprise Support, and 30% for Business Support+. Credits expire monthly. Queries to connected AWS services, such as CloudWatch Logs Insights, are billed separately at standard rates. Release Management is free during its preview. AWS DevOps Agent is also included in the AWS Free Tier free plan for new AWS customers.
Technical specs
API pricing
Same $0.0083 per agent-second; AWS CLI, SDKs, and CloudFormation, CDK, and Terraform support
Where AWS DevOps Agent excels
Overnight incident first response
Investigations can start automatically from alerts, so a root-cause analysis and mitigation suggestions are ready when the on-call engineer joins.
Incident war room in Slack
Engineers @mention the agent in a private channel, and its findings, team context, and recommended actions stay in a single thread.
Preventing repeat incidents
The agent analyzes historical incident patterns and recommends fixes across observability, infrastructure, deployment pipelines, and resilience.
Frequently asked questions
What does AWS DevOps Agent do?
It investigates production incidents on its own, pulling data from observability, code, and incident tools to find likely root causes and recommend mitigations. It also analyzes past incidents to suggest preventive fixes and answers on-demand SRE questions in chat.
How much does AWS DevOps Agent cost?
It is billed at $0.0083 per agent-second for investigations, evaluations, and on-demand SRE tasks. New customers get a 2-month free trial that includes up to 10 agent spaces and monthly hours of investigations, evaluations, and chat. AWS Support plans include monthly credits (30% to 100% of the prior month's support charges, depending on the plan). Connected services such as CloudWatch Logs Insights are billed separately.
Does AWS DevOps Agent work outside AWS?
Yes. It investigates incidents in Azure workloads and, through MCP, in on-premises applications, and it integrates with third-party tools such as Datadog, Dynatrace, Grafana, New Relic, Splunk, GitHub, GitLab, Azure DevOps, PagerDuty, and ServiceNow.
Can I use AWS DevOps Agent from Slack?
Yes. With bidirectional Slack communication, announced September 11, 2026, you can @mention AWS DevOps Agent in a connected private Slack channel to start and manage investigations, and ask about AWS resources, metrics, alarms, deployments, and incident patterns in the thread.
Is AWS DevOps Agent the same as Kiro?
No. Kiro is AWS's spec-driven agentic coding tool (IDE, CLI, and cloud agent) for writing code. AWS DevOps Agent works on running systems: it investigates incidents, finds root causes, and recommends mitigations.
Integrations & fit
About AWS DevOps Agent
AWS DevOps Agent became generally available on March 31, 2026. It can also investigate incidents in Azure workloads and, through MCP, in on-premises applications. You create an Agent Space in the AWS Management Console, connect your tools, and let the agent start investigations automatically when alerts fire or run them on request from its web app. During an investigation it pulls from observability platforms (Amazon CloudWatch, Datadog, Dynatrace, Grafana, New Relic, Splunk), code and CI/CD systems (GitHub, GitLab, Azure DevOps), and incident tools (PagerDuty, ServiceNow, Slack), keeps an investigation journal, and delivers a root-cause analysis with recommended actions. With permission, it can also carry out directed actions, such as a mitigation plan. Actions that create or modify resources are disabled by default. A triage agent assesses severity and flags duplicates. Beyond single incidents, it looks across past incident patterns to recommend preventive fixes, and it handles on-demand SRE questions in chat. Learned and custom skills capture how your organization operates, custom agents can run on a schedule or from webhook events, and A2A and MCP servers let you add your own tools. The AWS CLI, SDKs, CloudFormation, CDK, and Terraform cover automation. In September 2026, AWS added bidirectional Slack communication, so engineers can run investigations by @mentioning the agent in a private Slack channel, with findings and team context kept in one thread. A Release Management capability, in preview and only in US East (N. Virginia), reviews changes for release readiness and generates change-specific test plans. Unlike the coding agents in this directory, it works on running systems rather than code. The tradeoffs are cost and scope. Pricing is $0.0083 per agent-second of work, which comes to about $30 per agent-hour, plus charges from connected AWS services such as CloudWatch Logs Insights. AWS Support plans offset part of that with monthly credits. Agent Spaces can be created in 11 AWS Regions, although the agent can investigate resources in any Region, and it works best for teams that already have observability and incident tooling in place.
Updates from AWS DevOps Agent
AWS DevOps Agent added bidirectional Slack communication. Engineers can @mention the agent in a connected private channel to start and manage investigations, with findings and recommended actions kept in one thread.
AWS DevOps Agent became generally available, adding Azure and on-premises investigations, PagerDuty, Grafana, Azure DevOps, and Amazon EventBridge integrations, and CLI and SDK support, with per-second pricing and AWS Support plan credits.
Are you the founder? Claim this listing →