fridaySRE

Lets have infrastructure monitoring directly into IDE using AI. Ask 'Why is my app slow?' and get instant answers. No dashboards, no context switching - just natural language DevOps

Comment

fridaySRE - AWS MCP Hackathon Submission

Inspiration

We watched DevOps engineers waste hours switching between terminals and dashboards during incidents. We thought: what if infrastructure monitoring could be as simple as asking a colleague for help? fridaySRE brings AI-powered SRE capabilities directly into your IDE, making every developer as effective as a senior DevOps engineer.

What it does

fridaySRE is an AI infrastructure assistant for Cursor IDE that lets you diagnose issues in plain English. Ask "Why is my pod crashing?" or "When did we have peak traffic?" and get instant answers. It monitors Kubernetes, analyzes metrics, learns from past incidents, and provides cost-saving recommendations—all through natural conversation.

Key features:

  • Natural language diagnosis with 19+ specialized tools
  • Real-time Kubernetes and Prometheus monitoring
  • Historical analysis using TigerData (TimescaleDB)
  • Incident pattern matching with Redis vector search
  • 10x faster root cause analysis through parallel execution

How we built it

We used the Model Context Protocol (MCP) to integrate with Cursor IDE, Python AsyncIO for parallel agent execution, TigerData for storing months of metrics, and Redis for lightning-fast incident search. Our multi-agent architecture runs specialized agents (Kubernetes, Prometheus, SLO, Log Analysis) concurrently, reducing diagnosis time from 15 minutes to under 2 minutes.

Challenges we ran into

  • Redis Authentication: Complex Kubernetes networking required custom port forwarding solutions
  • Database Types: PostgreSQL Decimal types broke JSON serialization, requiring custom encoders
  • Parallel Coordination: Managing concurrent agents while maintaining data consistency
  • PromQL Syntax: Complex metric queries needed careful handling of time ranges

Accomplishments that we're proud of

  • 87% success rate with 20/23 tools working perfectly
  • 10x faster incident diagnosis (15 min → 2 min)
  • 90% storage reduction while maintaining sub-second query speeds
  • Natural language interface that makes SRE accessible to all developers
  • Already preventing outages in production environments

What we learned

Parallel execution isn't just faster—it enables correlation analysis impossible with sequential tools. Combining real-time monitoring with historical data reveals invisible patterns. Most importantly, natural language interfaces democratize DevOps expertise across entire teams.

What's next for fridaySRE

  • Automated remediation with safety checks
  • Predictive alerting to prevent incidents before they happen
  • Multi-cloud support and GitOps integration
  • AI-generated runbooks from incident patterns
  • Cost optimization with 20-40% typical savings

fridaySRE transforms infrastructure monitoring from reactive firefighting to proactive conversation. By making SRE as simple as asking a question, we're revolutionizing how developers interact with their infrastructure.

Built for AWS MCP Hackathon using TigerData, Redis, and the Model Context Protocol.

Built With

Share this project:

Updates