AWS DevOps Agent
Also known as: Amazon Web Services DevOps Agent
AWS DevOps Agent investigates incidents and proposes mitigations for teams operating AWS, other clouds and on premises systems. It connects to monitoring, ticketing and source control systems. An Agent Space defines which resources and tools the agent can use. The service can start from an alert, investigate in parallel and keep a timeline of its findings. Operators can review actions, correct root causes and replay previous investigations.
AWS DevOps Agent is an Amazon Web Services product for incident response and operational reliability. An alert, a ticket or a manual request can start an investigation. The service compares new incidents with ongoing work, then links related events or starts a separate case. An investigation can assign subagents to gather signals in parallel, such as metrics for different services. The agent produces a root cause analysis, a summary of evidence and possible mitigations.
An operator can send the running investigation additional facts or ask it to reconsider a hypothesis. Between incidents, the agent studies past ones and recommends improvements to observability, infrastructure, deployment pipelines and resilience, which can go to a coding agent as specs, and on demand it answers questions about resource health and deployments or runs a scheduled task such as a daily database health report.
An Agent Space defines the accounts, resources and tools accessible to the investigation. AWS DevOps Agent brings together deployment, observability and infrastructure relationships in a topology that can link a running resource to its code. It can connect to ServiceNow, Datadog, Grafana, GitHub and other systems. Administrators can connect MCP tools and designate which can change data. An elevated tool action can wait while a person approves or rejects it. The timeline preserves nested steps and tool activity, and a completed investigation can be replayed with changes to skills, instructions or integrations and compared with the earlier run.
The agent keeps managed memories about monitored alarms, earlier investigations and feedback. A monitor memory expires 14 days after the last update. If an operator corrects a root cause, subsequent investigations can use that correction. Agent Spaces can be created in supported AWS Regions, and customer data is stored in the region selected for the space. Organizations can use IAM Identity Center for operator access. Operations teams can also start investigations through an API or connect other assistants through the AWS DevOps Agent MCP or Agent to Agent interfaces.
AWS bills active agent time for incident investigations, site reliability evaluations and on demand SRE tasks at $0.0083 per second. New DevOps Agent customers have a two month trial with separate monthly allowances. Connected AWS services can still incur their own charges. Release management, still in preview, reviews a change for release readiness, maps dependencies across repositories, checks infrastructure changes for permission drift and generates tests for web and API apps.
Western Governors University cut resolution time from about two hours to 28 minutes, a 77% improvement, and Zenchef investigates in 20 to 30 minutes instead of one to two hours. Audible resolves incidents in a median of 8.5 minutes with the service, Banco Bmg set up the agent to investigate more than 350 production incidents, and Deriv gets a safe, caution or block verdict on a release in minutes rather than hours. United Airlines, T-Mobile and TP ICAP also use it. A reliability team can integrate a ServiceNow queue, use the agent to document a suspected source of a spike and review which mitigations to take.
Vendor details
Canonical URL
https://aws.amazon.com/devops-agent/
Category
SRE / DevOps agent
Subcategory
Incident investigations and operational reliability for cloud teams
Company status
first party product
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Amazon CloudWatch, Datadog, Dynatrace, New Relic, Splunk, Grafana, ServiceNow, PagerDuty, Slack, GitHub, GitLab, Azure DevOps and custom MCP servers.
In practice
A ServiceNow ticket starts an investigation, and the agent sends the ticket updates with evidence and a possible mitigation.
A reliability engineer sees which subagents gathered metrics across services and asks the lead investigation to check another hypothesis.
A team changes an investigation skill and replays a resolved case to compare the new path and grade with the original.
Sources & related URLs
Research sources
Agentic Index coverage score
13.0 / 14 capabilities · 93%
| Integrations & Tool Calling | Full |
|---|---|
|
An Agent Space can connect outside observability, ticketing and source control systems. Its administrator can register an MCP server and select specific tools to make available during investigations. A tool can be classified for read access or for actions that change data. When an incident begins in ServiceNow, the agent can send results and mitigation proposals into the original ticket. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/configuring-integrations-and-knowledge-connecting-mcp-servers.htmlread 2026-11-20 |
|
| Workflow Orchestration | Full |
|
A new incident is evaluated against other active work to decide whether to link it to a current investigation or assign a separate case. The investigation can delegate specific work to multiple subagents, such as checking CPU metrics across separate services. It combines the findings and can surface mitigation proposals after forming its root cause analysis. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/production-operations-autonomous-incident-response.htmlread 2026-11-20 |
|
| Knowledge Grounding & RAG | Full |
|
The service maintains a topology of the applications and resources that an Agent Space can investigate. It combines resource information with observed behavior and deployment data. That map helps the agent connect a dependency or a deployed resource to a source repository and gives later investigations a way to follow the environment’s relationships as they collect evidence. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/about-aws-devops-agent-what-is-a-devops-agent-topology.htmlread 2026-11-20 |
|
| Human Oversight & Guardrails | Full |
|
When an elevated action requires approval, the agent pauses at the specific proposed tool invocation. The service sends an approval request so an operator can accept or reject the action. That decision is associated with the paused execution, letting the investigation continue after the operator acts. Administrators can configure which external tools are treated as changing data. Sourcedocs.aws.amazon.com/devopsagent/latest/apireference/api_approvalaction.htmlread 2026-11-20 |
|
| Security, Identity & Governance | Full |
|
AWS IAM permissions control the accounts and resources that an Agent Space can reach. Administrators can grant operators access through IAM Identity Center and configure identities for external systems. The service encrypts stored customer data and supports customer managed keys. Third party AWS audit reports are available through AWS Artifact for the applicable service scope. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/aws-devops-agent-security.htmlread 2026-11-20 |
|
| Observability & Auditability | Full |
|
The investigation timeline shows steps the agent took, tools it called and when a step occurred. A subagent entry can be expanded to show the context and task assigned to it, its reasoning steps and its findings. An operator can follow which evidence and tool responses led to the result and how each delegated step contributed to the larger investigation. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/production-operations-autonomous-incident-response.htmlread 2026-11-20 |
|
| Memory & State Persistence | Full |
|
The agent keeps separate managed stores for topology, monitors, reflections and tool use lessons. A monitor memory can summarize recurring causes associated with an alarm and expires 14 days after its most recent update. A correction from an operator can create a reflection that later investigations use. Stored relationships help the agent locate a relevant lesson or operational fact. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/devops-agent-memories-managed-memories.htmlread 2026-11-20 |
|
| Deployment & Data Residency | Full |
|
A buyer chooses a supported AWS Region for an Agent Space and can create spaces in different locations. Production operations and on demand tasks are available in supported regions, including a GovCloud option subject to its feature and connection limits. A space can investigate authorized resources across AWS Regions while keeping the service’s stored data in the space’s selected region. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/about-aws-devops-agent-supported-regions.htmlread 2026-11-20 |
|
| Prebuilt Agents / Templates / Packs | Full |
|
An operator can add a published skill from the Skills Catalog to an Agent Space. A skill supplies reusable operational instructions for a particular kind of task and can target incident triage, root cause analysis or mitigation. The agent loads relevant active skills during the investigation. Customers can also add their own skills from a repository or archive. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/about-aws-devops-agent-devops-agent-skills-index.htmlread 2026-11-20 |
|
| Triggers & Channel Coverage | Full |
|
Supported integrations can start an investigation from a ticket or event. ServiceNow can start one for a ticket, and a webhook can send an alarm event from Grafana or a ticket from PagerDuty. The console also supports manually starting a case from a description or a selected alarm. The triggering data gives the agent a starting point for the investigation. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/production-operations-autonomous-incident-response.htmlread 2026-11-20 |
|
| Model Flexibility & Routing | Partial |
|
Amazon Bedrock foundation models power AWS DevOps Agent. Administrators can configure the agent’s instructions, accessible tools and knowledge sources within an Agent Space. The operational workflow uses AWS managed model access. A reliability team can change its investigation skills and compare the result of the changed configuration using the service’s replay surface. Sourceaws.amazon.com/devops-agent/faqsread 2026-11-20 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
The AWS DevOps Agent API lets external software manage Agent Spaces, start investigations and retrieve findings. Buyers can also connect another assistant through an MCP interface or use the Agent to Agent endpoint with supported credentials. That endpoint supports messages, streamed responses and asynchronous work. Webhooks allow an outside alerting service to start an investigation. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/working-with-devops-agent-accessing-devops-agent-index.htmlread 2026-11-20 |
|
| Testing, Debugging & Optimization | Full |
|
An operations engineer can replay a closed investigation after changing a skill, an instruction or an integration. The web application lets the engineer compare the replay and the original by grader score, duration and timeline. Replays run against available operational data and leave the original investigation intact. The operator can compare multiple replays of the same incident. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/working-with-devops-agent-replaying-investigations.htmlread 2026-11-20 |
|
| Browser / Computer-use | Partial |
|
An investigation timeline can show the browser steps used when the agent authenticates to a downstream system and encounters an error. This gives an operator detail about what occurred during a live investigation. The release management preview has a separate web application testing capability that generates plans tailored to a code change and executes them in the customer’s test environment. Sourcedocs.aws.amazon.com/devopsagent/latest/userguide/production-operations-autonomous-incident-response.htmlread 2026-11-20 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
$0.0083 per active agent second
Active agent seconds and separate charges from connected services
Included quota
Trial includes up to 10 Agent Spaces and, each month, 20 investigation hours, 15 evaluation hours and 20 on demand SRE hours.
What is public
Investigations, site reliability evaluations and on demand SRE tasks cost $0.0083 per active agent second. The service charges for work time. Release management has no additional charge during preview.
Billing mechanics
AWS meters cumulative active agent time per second. Connected service charges are billed at standard rates. Trial allowances apply separately to investigation, evaluation and on demand hours.
Variable cost rationale
The bill rises with the active time used across operational tasks. Connected AWS services such as CloudWatch Logs Insights may charge separately.
Overage / add-ons
Investigation, evaluation and on demand tasks cost $0.0083 per active agent second outside the applicable allowance.
Sales call required
No, self serve available
Free / trial
Two month trial for new DevOps Agent customers
Lowest paid plan
AWS DevOps Agent operations
Commercial notes
Eligible paid AWS Support customers receive monthly credits at 100% of the prior charge for Unified Operations, 75% for Enterprise Support and 30% for Business Support+.
Missing data
A buyer’s active agent time and charges for connected services depend on the workload.
Related vendors
- AlertD — AI agents for AWS operations that run read only in the customer's…
- Anyshift — AI SRE built on a versioned knowledge graph of infrastructure, apps…
- Avesha — Obliq, Avesha's autonomous AI SRE for Kubernetes and agentic…
- Better Stack — Better Stack offers incident management with built in on call,…
- BigPanda — BigPanda is an agentic IT operations platform whose Biggy agents…
- Bluebricks — Context and control layer that lets AI agents operate cloud…
Alternatives to AWS DevOps Agent
Other products that do the work AWS DevOps Agent does, and what each one is.