AWS DevOps Agent reaches GA with PagerDuty, Datadog and LaunchDarkly integrations

AWS announced general availability of its DevOps Agent, an agentic system aimed at autonomous production-incident resolution. The headline integration pairs the agent with Datadog's MCP Server to correlate monitoring data with deployed infrastructure and resolve incidents in minutes rather than hours. AWS rolled out a wave of complementary connectors: a PagerDuty integration to speed root-cause analysis for SRE teams (cutting the time spent triaging '2 a.m. alerts' by explaining why something broke, not just that it broke), and a LaunchDarkly integration that identifies relevant feature flags and coordinates disabling them across teams during outages.

Mechanically, the agent leans on the Model Context Protocol to plug into the observability and incident-management stack engineers already run, then reasons over traces, metrics, and topology. AWS also introduced 'Kiro power' for the DevOps Agent, surfacing cloud metrics, traces, topology and configurations directly in the IDE so engineers get cloud intelligence next to their code during triage.

Competitively, the GA push is part of AWS's broad agentic offensive this week — alongside Web Search on Bedrock AgentCore (GA), the Continuum security platform, and Amazon Quick autonomous agents. AWS is positioning agentic ops as the killer enterprise use case where reliability and integration matter more than raw model intelligence, directly countering Microsoft's Copilot and Google's orchestration story.

The caveat is the week's recurring reliability theme: a Martin Fowler HN piece (186 pts) on 'building reliable agentic AI systems' and dev.to posts dissecting how multi-step pipelines compound failure rates underscore skepticism about 'autonomous resolution in minutes' claims. Watch next: real-world MTTR data and whether teams trust the agent to act, not just advise.