AWS DevOps Agent: an "on-call engineer" agent for incidents

2025-12-02 · Publisher : AWS · Domain : Cloud & DevOps

On 2 December 2025, AWS launches in public preview AWS DevOps Agent, an agent that correlates metrics, logs and deployments to find the cause of an incident. It has been generally available since 31 March 2026.

### Background On 2 December 2025, AWS announces the public preview of AWS DevOps Agent, described as a "frontier agent": autonomous and able to work for hours or days without constant intervention. It presents itself as an always-on on-call engineer that, during an incident, correlates data across the operational toolchain, identifies probable root causes and recommends mitigations. ### Key facts - Observability integrations: Amazon CloudWatch, Datadog, Dynatrace, New Relic and Splunk. - Deployment tracking through GitHub Actions and GitLab CI/CD; tickets through ServiceNow (built in) and PagerDuty (via webhooks). - Custom MCP servers supported to plug in other tools, such as Grafana or Prometheus. - The agent coordinates the incident in Slack and maintains an investigation timeline, from an application topology it builds itself. - Free preview with task-hour limits, available in the US East (N. Virginia) region; 31 March 2026 update: general availability, with a two-month free trial. ### Why it matters The agent illustrates AI arriving in operations, beyond coding assistance: it reads deployments, metrics and tickets to shorten root-cause analysis. For CIOs and consulting firms running production, the question is access: which data and permissions to give an agent whose recommendations will need checking. Its openness to MCP servers also shows operations tooling organizing around this protocol. ### Takeaway Hyperscalers make on-call a field for agents: operations teams must frame access, verification and accountability.

Tags : Agents, Observabilité, MCP, AWS, IA