AWS DevOps Agent can autonomously investigate many EKS production incidents, but its visibility is limited to native integrations. This walkthrough shows how to extend it by building a custom Model Context Protocol (MCP) server that exposes EKS worker node OS-level diagnostics — iptables rules, CNI config, kubelet logs, dmesg, and 20+ other sources — via AWS Systems Manager Automation. The MCP server is deployed with AWS CDK, exposed through Amazon Bedrock AgentCore Gateway with OAuth authentication, and connected to AWS DevOps Agent without giving the agent direct shell access. A concrete demo injects a silent DNS-breaking iptables fault on a node (pods stay Running but DNS fails), and the agent autonomously identifies the root cause by chaining collect, triage, comparison, and correlation tool calls. The same pattern generalizes to any data source unreachable by the agent's native integrations.

14m read timeFrom aws.amazon.com
Post cover image
Table of contents
PrerequisitesHow AWS DevOps Agent discovers custom tools through MCPWhy Amazon EKS node OS visibility mattersIntegrating AWS DevOps Agent with an EKS node diagnostics MCP serverSeeing it in actionExtending this pattern to other data sourcesClean upConclusionAbout the author
136 Impressions