Tmap Mobility Automates Identification of 77% of Incident Root Causes with AWS DevOps Agent

Tmap Mobility built an AIOps system by integrating its in-house agent Rookie with AWS DevOps Agent.
Conducting 526 in-depth investigations over 7 weeks, it derived results in an average of 17 minutes and secured reviewable root cause hypotheses for 77% of all requests.
Leveraging the Memories feature and Learning agent of AWS DevOps Agent, it automatically accumulated past investigation knowledge and automatically merged duplicate alarms into existing incidents.
It enhanced investigation quality by integrating in-house systems such as Argo CD and Datadog via Custom MCP and applying the AWS open-source eks-troubleshooting Skill to track infrastructure change history.

Adoption Background and Architecture

The in-house developed agent Rookie required further verification in 68.4% of cases during primary analysis, so the in-depth investigation role was separated to AWS DevOps Agent.
Rookie receives requests from Slack and Jira, delegates investigations to DevOps Agent via Frontier (Custom MCP), and analyzes in parallel with Datadog Bits AI and DB DOSA.
DevOps Agent automatically identifies AWS resource dependencies through Agent Space and accesses MCP servers in private subnets using VPC Lattice.

Applying the eks-troubleshooting Skill from the AWS official open-source sample-ai-agent-skills, it accurately identified root causes of incidents caused by Karpenter node optimization tasks.
Integrating Argo CD as a Custom MCP tool allowed the agent to directly query service deployment locations, resource trees, and recent Revision metadata.
Specifying standard runbooks in the Skill provided inference guidelines for cross-verifying deployment resources, change management, and service dependencies.