December 2025
MakanMana
Your AI-powered food plug for the best makan spots in Malaysia. Ask naturally in Manglish and it hooks you up with real places, backed by real Google Maps reviews. Built with Syahirah for Cursor Hackathon 2025.
A hackathon build with Syahirah — we wanted food discovery that talks like your friends do ("eh nak makan sedap kat mana?"), not another filter-and-sort directory.
MakanMana is your AI-powered food plug for discovering the best makan spots in Malaysia 🇲🇾. You just ask naturally in Manglish — "eh nak makan sedap kat Bangsar?" — and it hooks you up with authentic spots, backed by real Google Maps reviews. No cap. 🔥
What it does
- 🤖 Conversational recommendations — chat in Manglish, powered by Google Gemini 2.0 Flash.
- 📍 Location-based search — find spots near where you actually are.
- 🔍 Real Google reviews — pulled live via SerpAPI, not made up.
- 💰 Budget & cuisine filters — nasi lemak on a student budget, sorted.
- ⭐ AI "vibe check" — a quick read on the atmosphere of each place.
- 🖼️ AI-generated food images and Manglish comments for flavour.
- 🔖 Save your favourites, 🌙 dark mode, and a fully responsive layout.
How it's built
- Frontend — React + TypeScript + Vite + Tailwind CSS, with shadcn/ui components.
- Backend — Supabase (PostgreSQL + Edge Functions), Row Level Security on every table.
- AI — Google Gemini 2.0 Flash for chat, vibe analysis, food images, and comment generation.
- Data — SerpAPI for Google Maps search & reviews.
Four Supabase Edge Functions carry the load: chat (the assistant), vibecheck (restaurant vibe analysis), generate-food-image, and generate-comment (the Manglish flavour).
The team
Built together at Cursor Hackathon 2025 🇲🇾:
- Amirah — Supabase Edge Functions, Google Gemini integration, SerpAPI / Google Maps integration, and AI prompt engineering.
- Syahirah — Frontend & AI Groq integration, React application architecture, UI/UX design, and responsive theming.
Design inspired by Malaysian hawker centres — Sambal Red, Pandan Green, Nasi Cream, and friendly Poppins type. Made with ❤️ in Malaysia.
More projects
view all →
Kaeru Desktop App
A lightweight AI workspace anyone can use: chat with an agent to build and edit, preview your app live, and keep your work safe with plain-language source control. A desktop app built on Tauri + Rust, about 8 MB.
Canopy
An agent situation-room: a panel of AI leads debates a call across your projects, then you approve it. Bounded, mediated, human-approved.
Kaeru UI harness
A case study on how Kaeru's interface gets designed: one page where every screen and state is mocked, reviewed, and promoted from design mockup to real component — before any of it ships.