← All work

MoutMessenger

GitHub

Different clients. One conversation.

RoleAI Product Developer
StatusWorking product · self-hosted
StackReact 19 / TypeScript / Electron / Express / Socket.io / Prisma / PostgreSQL
WEB
DESKTOP
REALTIMESocket.io
Architectural composition — not a product screenshot

Conversations. In real time.

A unified backend for web and Electron clients, with realtime delivery and operational tooling.

The server validates chat membership and message permissions before broadcast. Electron adds desktop capabilities through a preload bridge. Storage protection is server-side, not end-to-end encryption.

01

React / Electron

Shared interface, multiple clients

02

Express + Socket.io

Authorization and realtime delivery

03

Prisma + PostgreSQL

Persistent product data

What is verified.

Implemented

  • Private and group chats, reactions, replies, forwards, editing, and read receipts.
  • Socket.io delivery, typing, presence, media, stories, call signaling, and admin/moderation flows.
  • Desktop: tray, notifications, background sync, local cache, and a narrow preload bridge.

Basis for the claims

npm workspaces: web, server, desktop, and a mobile foundation.

Express + Socket.io + Prisma architecture, Electron preload bridges, and deployment scripts.

The security model is server-authoritative. Optional server-side storage protection exists; E2EE is not claimed.

Mout AI