kiro-haunting.me

Inspiration

This project was a last-minute creation based on my learnings from aws:Reinvent 2025. Excited by the progress that has been made in the Generative LLM space, I set out to challenge myself to produce a fully-shipped product in under 24 hours. I've participated in many hackathons before and feel I have a good understanding of what used to be possible for a single skilled dev in that timeframe.

But not anymore - Kiro Code completely transforms what is possible in that timeframe.

What it does

kiro-haunting.me transforms ordinary smart homes into dynamic haunted house experiences using AI-powered orchestration. The application connects to Amazon Alexa or Google Home ecosystems to discover and control smart devices (lights, speakers, TVs, smart plugs), then deploys specialized AI sub-agents to create coordinated spooky behaviors.

Multi-Agent Architecture: Main orchestrator coordinates specialized sub-agents for each device type (lights, audio, TV, smart plugs)
Two Operation Modes: Connected mode (direct device control) or Simple mode (voice commands via browser)
Customizable Themes: Multiple haunting styles from "Subtle Unease" to "Poltergeist"
Safety Features: Epilepsy-safe mode prevents dangerous strobe effects
Real-time Monitoring: Live activity feed shows what each agent is doing
Halloween UI: Spooky themed interface with atmospheric effects

How we built it

This codebase and infrastructure was built almost entirely with Kiro code, leveraging MCP servers for documentation and AWS knowledge.

Tech Stack: Vue frontend, AWS Lambda backend, OpenAI-powered agents, DynamoDB storage, fully serverless architecture.

Challenges we ran into

Learning the AWS stack and familiarizing myself with the best context-engineering practices to steer the Kiro Code agents were the primary hurdles I crossed to complete this project.

Accomplishments that we're proud of

With almost zero prior understanding of the AWS stack, I shipped a production grade cloud deployment with a fully featured frontend and backend, live to https://kiro-haunting.me in under 24 hours from start to finish - with only about 8 hours of actual dev time.

What we learned

Generative AI has completely changed what is possible and will radically shift the workload of talented software developers away from menial labor tasks such as writing for-loops, constructing unit tests, and tinkering with devops infrastructure and allow a creative developer to focus on their biggest value add: Architecture. It's a force-multiplier like we've never seen before.