What is AWS re:Invent?
re:Invent is AWS’s biggest event of the year. It runs every late November in Las Vegas, and it is where AWS ships most of its headline launches: new services, big feature drops, and whatever direction the platform is heading next. Between the keynotes and hundreds of product announcements, it is the week most of us watching the cloud space actually pay attention to.
What AWS Announced
AWS put DevOps Agent into public preview. It is a frontier agent that investigates incidents on its own, works out the root cause, and then suggests changes meant to keep the same thing from happening again. The pitch is that it behaves less like a monitoring dashboard and more like an extra engineer on the team, one that reads your telemetry, code changes, deployments, and operational patterns across the environment without being asked twice.
During preview it costs nothing extra. It runs in US East (N. Virginia), and it can watch applications deployed in any Region and across multiple AWS accounts.
Why This Matters
If you have ever been paged at 2am, you know the drill. You open five tabs: metrics in one, logs in another, the deploy history, the dashboards, the dependency map. You stitch the story together by hand while the clock runs and someone in Slack asks for an ETA. The tooling is everywhere and connected to nothing, and that fragmentation is usually what slows root cause analysis down, not the actual problem.
Then the incident closes, everyone goes back to their backlog, and nobody circles back to ask why it happened in the first place.
DevOps Agent is aimed squarely at that gap. It runs the investigation for you, pulls correlated signals from across the toolchain instead of one silo at a time, and hands back recommendations meant to harden the system. Less time spent stitching tabs together means lower mean time to resolution, and the pattern analysis is what finally makes the post-incident review actually happen.
Key Capabilities
Autonomous Incident Response
The agent starts digging the moment an alert lands, whether it comes from CloudWatch or an external tool like ServiceNow or PagerDuty. From there it correlates metrics, logs, traces, and recent deployments pulled from GitHub, GitLab, Datadog, Dynatrace, New Relic, and Splunk.
It surfaces the likely root causes and posts findings, a timeline, and recommendations straight into the incident channel in Slack. It will also run point as a virtual incident coordinator, handling the communication and stakeholder updates that normally eat an on-call engineer’s attention.
Interactive Investigations
You are not locked out of the loop. There is a web app where on-call teams can kick off investigations, dig into the analysis, look at the application topology, and ask follow-up questions in plain language. If the agent is chasing the wrong thread, you can feed it more context, narrow the scope, or point it at specific resources or logs.
Proactive Operational Improvements
DevOps Agent does not stop once the fire is out. It looks across past incidents to find the systemic gaps, then gives you specific recommendations on observability, infrastructure configuration, capacity tuning, testing, and the quality of your deployment pipeline. That is the difference between firefighting and reliability engineering.
Intelligent Application Topology
The agent keeps a live topology graph of your AWS resources and how they connect: compute, storage, networking, and deployment history. That map is what lets it reason about how a change in one place ripples into another while an investigation is underway.
Extensible Tool Integrations
You can wire in more tools through the Model Context Protocol (MCP), so the agent can pull from open source stacks like Prometheus and Grafana or your own internal tooling. The point is one investigation surface instead of a dozen, even when your environment spans multiple clouds and hybrid setups.
How It Works
Work is organized into Agent Spaces. These are containers that spell out which AWS accounts, tools, and resources a given agent is allowed to touch. Admins set up Agent Spaces and integrations in the AWS Management Console; operators live in the web app for the day-to-day investigations.
Investigations kick off automatically when an alarm fires, or you can start one by hand from a predefined path: recent alarm triggers, high CPU, a spike in application errors, that sort of thing.
Once it is running, the agent works through your telemetry, traces the relationships in your application stack, reviews recent deployments, and writes it all up in a single summary with root cause candidates and mitigation guidance.
Availability and Preview Details
DevOps Agent is in preview now in US East (N. Virginia). The agent runs in us-east-1, but it monitors workloads in any Region and across multiple accounts. Preview is free, with one caveat: there is a monthly cap on agent task hours.
Sign up for the preview and start connecting your observability and deployment tools from the AWS Management Console.
If you want to put this to work on real incident response and tighten up how your systems hold under load, our cloud experts can help you fold AWS DevOps Agent into an operations setup that scales with you.