SRE-AGENTIC

Inspiration

We used sentry on an internship, wanted to see how much it could help improve the reliability of our platform.

What it does

It saves time by helping engineers understand what's happening beforehand.

How we built it

Built a demo webpage to deliberately have errors, let Sentry log them, and then we sent the alerts to our agents to diagnose via python tools ran on Daytona. It uses Twilio's API to make phone calls to engineers with an initial diagnosis.

Challenges we ran into

Using Twilio's API sucked, so we get the phone call, but Twilio doesn't allow us to send the voice message with the diagnosis report.

Accomplishments that we're proud of

Calls work :) and so does the whole infrastructure.

What we learned

We can automate very different tasks with this. We can add logs from an SOC to automate incident response, let the agent recognize bugs, solve them and push them to production autonomously, and even we could analyze the structure of the team to precisely route the problem to the right person.

What's next for SRE-AGENTIC

Winning.

For AI

## Inspiration

During an internship, we relied heavily on Sentry for error monitoring and realized its potential went far beyond just logging—what if we could use it to proactively improve platform reliability by combining it with AI-powered diagnostics?

## What it does

SRE-AGENTIC reduces incident response time by automatically diagnosing errors the moment they occur. When Sentry detects an issue, our system analyzes it and calls the on-call engineer with a preliminary diagnosis—so they understand the problem before they even open their laptop.

## How we built it

• Demo environment: A web application designed to generate realistic errors for testing
  
• Error pipeline: Sentry captures errors and forwards alerts via webhooks to our agent system
  
• AI diagnostics: Agents analyze the error context using Python tools running on Daytona
  
• Engineer notification: Twilio API initiates phone calls to alert engineers with the diagnosis
  

## Challenges we ran into

Twilio's API proved frustrating—we successfully initiate calls to engineers, but limitations prevented us from delivering the full voice-based
diagnosis report as intended.

## Accomplishments that we're proud of

• End-to-end pipeline works: error → diagnosis → phone call
  
• Successfully integrated Sentry, Daytona, and Twilio into a cohesive system
  
• Demonstrated real reduction in time-to-understanding for incidents
  

## What we learned

This architecture opens doors to automating far more than error alerts:

• SOC integration: Feed security logs to automate incident response
  
• Autonomous bug fixes: Let agents identify bugs, implement fixes, and push to production
  
• Intelligent routing: Analyze team structure to route issues to the most relevant engineer
  

## What's next for SRE-AGENTIC

Expanding the agent's capabilities to not just diagnose, but remediate—moving toward truly autonomous site reliability engineerin

Built With

• typescript