Inspiration - JARVIS started as an Iron Man-inspired idea: an AI assistant that could see, hear, and act — not just answer questions. Over multiple versions, it grew into a real voice-driven system capable of automation, speech interaction, and even threat/phishing detection. When we saw the Alexa+ track, it felt like the natural next step: instead of keeping JARVIS confined to a desktop app, why not let it live inside the assistant millions of people already talk to every day? That’s where Bug404 was born — JARVIS’s intelligence, rebuilt to speak Alexa+’s language.

What it does - Bug404 is an MCP-powered Alexa+ skill that brings hands-free task automation and real-time threat detection to Alexa+. Users can simply talk to Alexa+, and Bug404 handles the rest — executing tasks and flagging potential threats, without needing a separate app or device.

How we built it - Bug404 is built on the existing JARVIS stack — Groq’s LLaMA 3.3-70b for reasoning, a dedicated automation layer, and a threat/phishing-detection module refined across earlier JARVIS versions (including JarvisShield, which placed Top 8/220 at EFOS SkillUp India). To make this accessible to Alexa+, we’re exposing these capabilities through a self-hosted MCP server (spec 2025-11-25, Streamable HTTP), so Alexa+ can call JARVIS’s core functions as tools in real time.

Challenges we ran into - Re-architecting a desktop-first voice assistant into a service Alexa+ can call via MCP Learning and implementing the MCP spec itself under a tight timeline Keeping threat-detection accuracy intact when triggered by voice commands instead of continuous background monitoring What we learned

We learned how MCP bridges custom AI logic with platforms like Alexa+, and how to modularize an existing assistant (JARVIS) so its core intelligence can be reused across completely different front-ends — from a PC app to a conversational voice skill.

Built With

Share this project:

Updates

Submission history