Jewel
All case studies
02Restaurant OS · Guest Memory

MenuMuze

The restaurant OS that remembers every guest.

Demosample data

Bookings, ordering, kitchen and bar, cashier, CRM and an AI host called M, all sharing one memory of each guest: allergies with their exceptions, diet and faith, favourites and the dates that matter. Guests say it once and the whole team knows.

1
guest profile, every screen
3
menu import paths: photo, PDF, text
Demo
fictional venue, real product
menumuze.getbijou.xyz
MenuMuze guest menu with M on a phone
React 19Express + WebSocketPostgres · DrizzleGeminiAgent APIWhatsApp via Bijou
01The problem

I spent years on restaurant floors as a waiter, mixologist, sommelier and then operations manager. The same facts get lost between stations every night: the guest's allergy and its one exception, the diet or faith rule, the anniversary they mentioned when they booked.

Restaurants run on five apps that don't talk: a booking app, a CRM nobody updates, POS add-ons per channel, paper tickets. At peak hours the person who knows is never the person serving.

02What I built
01

Guest Memory

One profile per guest, keyed to their phone number. Allergies store the form and the exception ("no pepper sauce, ground pepper is fine"), plus diet and faith, taste, occasions and staff notes.

02

Ask M

An AI host that answers from the venue's own dishes, knows who is asking, skips dishes that clash with their memory, and always asks staff to confirm allergens.

03

Back of house

Kitchen and bar displays per station with the guest's allergy on the ticket, a cashier with split bills, and staff alerts for new orders, "call waiter" and "the bill".

04

Front of house

A host view with tonight's timeline per table, late flags and M-drafted messages that the host reviews and sends.

05

CRM & loyalty

Segments that build themselves from guest memory, consent-aware re-invites drafted by M, and loyalty tiers earned only from bills closed at the cashier.

06

Menu import

Snap a photo, upload a PDF or paste text. M extracts dishes, prices, ingredients with their form, allergens and diet tags, and flags what it guessed for review.

07

Every channel, one pipeline

Dine-in QR, takeaway, delivery and online ordering read the same menu. WhatsApp ordering, powered by Bijou, and phone orders land in the same pipeline.

03Architecture & stack
Screens
Guest menu + MKitchen / barCashierHostCRMMenu Studio
App
React 19 + ViteTailwindZustand
Server
Node.js + ExpressWebSocket live updatesPer-venue tenancy
AI
Gemini for M and menu importAgent API for voice and WhatsApp
Data
Postgres via Drizzle ORM
04Screens

Captured from the live product, October 2026.

MenuMuze demo, guest menu with Ask M
Guest menu with Ask M (demo venue)
MenuMuze kitchen display
Kitchen display: tickets by station with allergy alerts
MenuMuze host view timeline
Host view: tonight's timeline and M's suggested messages
MenuMuze cashier with open bills
Cashier: open bills per table
MenuMuze guests and CRM
Guests & CRM: segments built from guest memory
MenuMuze menu import studio
Menu Studio: import from text, PDF or a photo
MenuMuze guest memory section
Guest memory: allergies with the detail that matters
MenuMuze on a phone
Guest view, phone
MenuMuze menu on a phone
Menu with fit badges, phone
05Honest notes
  • Status is Demo: Maison Obsidian and its guests are fictional, nothing is charged, and each demo sandbox resets after 48 hours.
  • The agent API is live. Connecting a venue's phone number and WhatsApp line is done per venue in a pilot, as is live payment processing.
  • There is no direct POS sync yet; it is scoped per venue.