AWS Builder Center
DevOps Sentinel
AI-Powered Incident Response

DevOps Sentinel AI-Powered Incident Response

DevOps Sentinel: AI powered AWS Incident response dashboard. simulates production incidents, investigates using metrics, logs, and MCP tools, then recommends Remediation with human Approval.

Building DevOps Sentinel – AI-Powered AWS Incident Response 🚀

Introduction

In a real production application, when something goes wrong, DevOps engineers have to check metrics, logs, deployments and AWS services to find what actually happend.
So I started building DevOps Sentinel, an AI-powered AWS incident response dashboard.
The main idea is simple:
What if AI can investigate a production incident, find the possible root cause and suggest what to do next?



💡 What is DevOps Sentinel?

DevOps Sentinel is a dashboard that simulates AWS production incidents and uses AI agents to investigate them.
The basic workflow is:
Incident
↓
Metrics & Logs
↓
AI Investigation
↓
Evidence Analysis
↓
Root Cause
↓
Recommendation
↓
Human Approval

Currently, the project runs in demo mode, so AWS credentials are not required.



🚨 Incident Simulator

One of the main features is the Production Incident Simulator.
With one click, it creates a fake production problem.
For example:
API Latency 180ms → 4.8s
Error Rate 0.4% → 12%
DB Connections 45% → 96%

It also generates logs showing database connection pool exhaustion.
This makes it easy to demonstrate the complete incident response flow without breaking a real AWS application 😅.



🧠 AI Investigation

The next part is the AI investigation engine.
Instead of just saying:
"Database is the problem."
The AI should collect evidence and explain why it thinks that.
Example:
Root Cause:
Database connection pool exhaustion

Confidence:
92%

Evidence:
- DB connections reached 96%
- API latency increased
- Error rate increased
- Connection timeout logs detected

This makes the AI decision more understandable and usefull.



🔌 MCP Integration

One of the main future features is a private MCP server.
The AI agent will be able to use Controlled DevOps tools such as:
get_cloudwatch_metrics()
get_application_logs()
get_database_metrics()
get_recent_deployments()

The idea is to give the AI only the tools it actually needs instead of giving it full AWS access.



☁️ AWS integration

The current version uses mock data.
The next phase will integrate real AWS services like:
  • CloudWatch
  • CloudWatch Logs
  • IAM
  • VPC
  • AWS CDK
  • AWS SDK
Security is also important here. I don't want the AI to have AdministratorAccess or unlimited permissions.
The plan is to use least-privilege IAM roles and human approval before important remediation actions.



🛠️ Tech Stack

The Project uses:
  • Next.js
  • React
  • TypeScript
  • Tailwind CSS
  • shadcn/ui
  • Recharts
  • AWS SDK
  • AWS CDK
  • Zod
I also created a provider architecture with a MockAgentProvider for demo mode and an AwsDevOpsAgentProvider for the future AWS version.



📈 Project Phases

Phase 1 – Dashboard ✅

  • Monitoring dashboard
  • Metrics charts
  • Incident simulator
  • Mock data

Phase 2 – AI Investigation 🚧

  • AI agent
  • Evidence collection
  • Root cause analysis
  • Recommendations

Phase 3 – MCP 🔌

  • Private MCP server
  • DevOps diagnostic tools
  • Tool call visualization

Phase 4 – AWS ☁️

  • CloudWatch integration
  • AWS CDK
  • IAM security
  • Human approval
  • Automated remediation



🚀 Running the Project

Clone the repo and install the dependencies:
npm install
npm run dev

Then open:
http://localhost:3000

No AWS credentials are needed for the current demo.



🎯 What I Learned

While building this project, I learned that building an AI agent is not just about connecting an LLM.
It also needs:
  • Proper tools
  • Good context
  • Evidence
  • Permissions
  • Security
  • Human control
Specially for DevOps, giving AI too much access can be risky.
So the goal of DevOps Sentinel is not to replace DevOps engineers. It's to help them investigate incidents faster and reduce the boring manual work.



🏁 Final Thoughts

DevOps Sentinel is still a work in progress and there is lot more to build, specially the MCP server and real AWS integration.
But this project gave me a good understanding of how AI + AWS + DevOps can work together.
I'm excited to continue building it and turn this simple demo into a more complete incident response platform. 🚀
GitHub:  github.com/vgplavan11-cmd/DevOps-Sentinel-AI-Powered-Incident-Response 
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article