Circle K brings AI investigation to incident alerts

case study describes an agent that investigates alerts using AWS resources, application logs, GitHub and service descriptions. Lambda starts its AgentCore session and sends the resulting summary to Slack. Gateway mediates access to data sources. Engineers receive findings and suggested fixes, then perform repairs themselves. Automated remediation with approval in Slack remains a plan. This is a company account, rather than an independent test.

Where the investigation runs

The Runtime documentation explains the hosting component: it runs agents and tools, including custom agents that use no particular framework. In its microVM mode, each user session has separate processor, memory and filesystem resources. Completing the session terminates that virtual machine and sanitizes memory. Built-in tracing records reasoning steps, tool calls and model interactions, giving developers information for debugging and auditing how an agent reached its output.

How tools become accessible

The Gateway documentation describes a shared entry point that turns APIs and Lambda functions into tools compatible with Model Context Protocol, a protocol connecting agents and tools. It distinguishes incoming authentication, which verifies the agent’s identity, from outgoing authentication used to connect to tools. The service handles OAuth flows, token refresh and credential storage. Contextual search also helps an agent find relevant tools within the available collection.