Transitly - Your AI-Powered Moving Concierge

💡 The Inspiration

Moving sucks. Not the physical part—the endless administrative nightmare that follows. After watching a friend spend 3 weeks manually updating their address across 47 different services (yes, they counted), losing $500 to missed utility deadlines, and still forgetting to forward their mail, we realized something had to change.

The question that started it all: What if AI agents could handle the bureaucracy of moving while you focus on actually, you know, moving?

That's when Transitly was born—an intelligent system that doesn't just remind you what to do, but actually does it for you.

🎯 What It Does

Transitly is a multi-agent AI system that autonomously handles moving-related tasks:

• 🤖 Intelligent Orchestration: A supervisor agent analyzes your move and delegates tasks to specialized agents
• 🌐 Browser Automation: Real agents that log into services like Amazon and update your address automatically
• 📋 Dynamic Planning: Generates personalized checklists based on your specific move timeline and locations
• ✅ Live Progress Tracking: Watch as AI agents complete tasks in real-time through our SSE-powered interface
• 🔒 Secure & Private: AWS Cognito authentication with encrypted session management

Input: Your old address, new address, and move dates
Output: A fully executed moving plan with addresses updated, services transferred, and nothing forgotten

🏗️ How We Built It

Architecture

┌─────────────────────────────────────────────────────┐
│  Frontend (React + TypeScript)                      │
│  • Real-time SSE streaming                          │
│  • Dynamic checklist UI                             │
└──────────────────┬──────────────────────────────────┘
                   │
┌──────────────────▼──────────────────────────────────┐
│  Backend (FastAPI + Python)                         │
│  • Multi-agent orchestration via LangGraph          │
│  • RESTful API + SSE endpoints                      │
└──────────────────┬──────────────────────────────────┘
                   │
        ┌──────────┼──────────┐
        │          │          │
   ┌────▼───┐ ┌───▼────┐ ┌──▼────────┐
   │ Gemini │ │  Nova  │ │ DynamoDB  │
   │  LLM   │ │  Act   │ │   AWS     │
   └────────┘ └────────┘ └───────────┘

Tech Stack

Frontend

• React 18 with TypeScript for type safety
• Vite for blazing-fast builds
• Server-Sent Events (SSE) for real-time agent status updates
• TailwindCSS for responsive design

Backend

• FastAPI for async Python web framework
• LangGraph for multi-agent workflow orchestration
• Google's Gemini 2.5 Flash for intelligent decision-making
• NovaAct for browser automation (Playwright under the hood)
• AWS Cognito for authentication
• DynamoDB for scalable, serverless data storage

Infrastructure

• AWS EC2 (c7i-flex.large) for compute
• Nginx as reverse proxy
• PM2 for process management
• Complete deployment on AWS

The Agent Architecture

Our system uses a supervisor-worker pattern with LangGraph:

1. Orchestrator Agent: Analyzes state and routes to appropriate specialist
2. User Details Agent: Fetches user profile and move information from DynamoDB
3. Checklist Agent: Generates personalized task list using LLM reasoning
4. Amazon Address Agent: Browser automation to update shipping address
5. (Extensible) Future agents for USPS, utilities, subscriptions, etc.

Each agent is idempotent and stateful—they check if work is already done before acting, preventing duplicate updates.

Built With

• cognito
• dynamodb
• fastapi
• langgraph
• nova
• python
• react