Rabbita AI

RabbitaAI is an autonomous AI agent that books any local service for you through WhatsApp in a single conversation , in any language, in any market , so you never have to make a search, text and deal

Comment

Inspiration

We've wanted this for years. AI booking agents keep appearing in the market, but they're either not practical or require lot of setup , which a lay man cant do. RabbitaAI is the first practical end-to-end solution. It actually works. No complex integrations, no waiting .. just describe what you need and it's booked. Nothing else on the market does this this simply.

What it does

You describe any service need in natural language (English, Urdu, Arabic, 40+ languages). The AI agent autonomously finds nearby providers via Google Maps, messages them on WhatsApp, negotiates details, and confirms the booking , all without you making a single call.

How we built it

Brain: Gemini on Google Cloud Vertex AI (true autonomous agent, not a chatbot) Data layer: MongoDB Atlas with MCP server (agent reads full context before every decision) Messaging: Meta WhatsApp Cloud API with request ID routing (solves multi-user routing at scale) Backend: Node.js/Express on Cloud Run Frontend: React/Vite on Firebase Hosting Real-time: Socket.io for live booking updates Auth: Firebase Authentication

Challenges we ran into

Maintaining conversation context across asynchronous WhatsApp exchanges Building genuine human-in-the-loop decision system (AI knowing what to handle vs. what needs user input) Multi-user message routing through a single WhatsApp business number without collision Mixed-language understanding (eg. Roman Urdu + Arabic + English in one message)

Accomplishments we're proud of

Live end-to-end bookings working in production MongoDB MCP integration that lets Gemini reason about data instead of executing raw queries Request ID prefix system that routes multi-user messages flawlessly Human-in-the-loop interface that surfaces only genuine decisions to users 40+ language support including colloquial/informal inputs

What we learned

Building a real AI agent means solving the problems around the AI, not the AI itself. The hardest work was conversation context, human-in-the-loop architecture, and routing , not reasoning. The MongoDB MCP shift from imperative to declarative data access fundamentally changed how the agent feels: genuinely intelligent instead of scripted.

What's next for RabbitaAI

Dominate the 2B WhatsApp user market. Expand to every city in Pakistan, Nigeria, Indonesia, Egypt, and beyond. Partner with local service ecosystems. Build the unified AI service layer that WhatsApp-first markets have always needed. Scale from hackathon to unicorn.

Built With

Share this project:

Updates