GuardScale: Your Infrastructure, One Reply Away
We started with a real production incident: automated traffic was overwhelming our website, and understanding why meant jumping between logs, metrics, and monitoring tools.
We used AI to investigate. Then we built GuardScale to bring the evidence, analysis, and response into one workflow.
And yes, the name changed. We misspelled the domain when buying the original Guardian-US name. That naming detour led us to GuardScale—a better fit for what we were building.
Communication became the backbone
We integrated AWS End User Messaging Social for WhatsApp and Amazon SES for email, making communication part of the response—not just an alert at the end.
Detect → Investigate → Ask → Protect → Verify
A lightweight Go agent watches server activity. Amazon Bedrock can request a closer inspection, assess the evidence, and recommend a response.
The operator reviews the findings and decides what happens next.
Unusual server activity detected. Review the observed request paths. Reply Protect to request temporary protection. The server agent confirms when the rule is active.
A working loop, with humans in control
Our final demo shows a disposable local server connected to the live GuardScale platform: suspicious requests generate an incident, an operator replies through WhatsApp, and the agent applies and confirms a temporary, scoped block.
The conversation continues with protection updates and expiry notifications.
We also expose controlled investigation capabilities through MCP, so other AI agents can use the same tools without bypassing human approval.
Explore what we built
Explore the code, architecture, and judge walkthrough
The goal remains simple: help teams understand what is happening, respond quickly, and keep humans in control.
What would you want to investigate or approve directly from a WhatsApp message?
Log in or sign up for Devpost to join the conversation.