MoveOS

Inspiration

Last summer, one of us landed an internship in Chicago and spent weeks drowning in Zillow tabs, Reddit threads, commute checks, and unanswered landlord emails all manually, all at once. It was exhausting. When we sat down to brainstorm ideas for this hackathon, that experience came up immediately. We knew exactly the problem we wanted to solve.

## What it does

MoveOS is an agentic AI that handles your entire relocation workflow in one place. You describe your move destination,workplace, budget, priorities and it takes over: discovering listings across Zillow, Craigslist, Reddit, RentCast, and Google Places; checking commute times for every candidate; scoring and ranking them; and drafting outreach messages to landlords. The whole pipeline runs autonomously, surfacing a ranked decision board so you can focus on choosing not hunting.

## How we built it

Google ADK — orchestrates the multi-agent pipeline end to end
Gemini — powers the AI reasoning, summarization, and outreach drafting
MCP servers — MongoDB MCP for persistent workspace storage, custom MCP tools for each data provider
FastAPI — async Python backend running the 9-stage discovery pipeline
Apify — Zillow listing scraper via actor API
RentCast API — rental market data and listing enrichment
Reddit API — community-sourced neighborhood recommendations
Google Places API — commute time validation and local context
Playwright — browser automation for Craigslist and other non-API sources
React + Vite + Tailwind — frontend with a Claude-style live agent workflow in chat
Firebase — authentication and user session management
MongoDB Atlas — relocation profiles, listing history, and outreach drafts

## Challenges we ran into

Getting heterogeneous data sources to normalize cleanly was harder than expected — each provider has different schemas, rate limits, and failure modes. Coordinating a 9-stage async pipeline (discover → enrich → deduplicate → commute-check → rank) without stalling or losing results required significant orchestration work. We also hit Apify quota issues mid-build and had to rearchitect the provider fallback logic under deadline.

## Accomplishments that we're proud of

A fully autonomous relocation pipeline that actually works end to end drop in a city and a workplace address, and within minutes you have a ranked, commute-validated shortlist with outreach drafts ready to send. The live agent workflow in the chat UI showing every step as it happens.

What we learned

Agentic pipelines live and die by their error handling. A single broken provider can silently kill half your results ifyou're not careful. We also learned how powerful MCP servers are as a composition layer swapping data sources in and out without touching the core agent logic was a huge win.

## What's next for MoveOS Expand to roommate matching, lease review via AI, and neighborhood safety scoring. Add email/SMS integration so outreach actually gets sent. Long-term: make MoveOS the default tool anyone uses the moment they accept a job offer in a new city.