WhatsApp on your desktop, built for the AI you choose.
WhatsZapp is an AI-native desktop client for Linux and Windows. Every message syncs into a local SQLite database on your machine. Your scripts, agents, and models can read that data today. Meta's assistant cannot.
Meta's AI lives inside their app. It suggests replies, pushes bots into your inbox, and runs on their servers with their rules. You cannot point it at a local model, pipe your full chat history into your own stack, or let an agent act on your inbox without sending everything through Meta first.
WhatsZapp flips that:
| What you want | Meta AI | WhatsZapp |
|---|---|---|
| AI on your terms | Their model, their UI | Bring your own model and API keys (roadmap) |
| Full history for agents | Locked in the app | Plain SQLite you can query with SQL today |
| Automate replies and workflows | Limited, inside WhatsApp | Agent API, webhooks, CLI (roadmap) |
| Privacy for sensitive chats | Processed on Meta's stack | Data stays on disk; only WhatsApp servers for sync |
| No unsolicited AI in the UI | Bots, banners, suggested replies | Chat first. AI when you wire it up |
WhatsZapp does not ship a chatbot in your sidebar. It ships the foundation Meta will never give you: a real desktop app plus a database your tools can actually use.
Today: Every message lands in ~/.config/whatsapp-desktop/whatsapp.db. WAL mode, fast sync, one file to back up or copy. You can already inspect chats with SQL, build dashboards, or feed exports into whatever model stack you run.
Coming (see Roadmap):
- Bring-your-own AI — Ollama, OpenAI, Anthropic, or whatever you run. Your keys, your prompts, your policy.
- Agent API — Local HTTP / MCP so agents can search threads, summarize groups, and draft replies without a browser tab.
- Smart inbox — Rules and filters you define (priority contacts, mute patterns, auto-labels).
- Programmatic send — Reply from scripts, cron, or your backend.
- Webhooks — Trigger automations when a message arrives.
- Reply from the terminal — One command to answer without opening the window.
That is the gap Meta AI cannot close: they optimize for engagement inside WhatsApp. WhatsZapp optimizes for your automation stack outside it.
Linux and Windows get a proper Electron window: system theme, split inbox inspired by macOS Messages, tray-friendly workflow (tray icon on the roadmap). No tab lost among twenty others, no "please keep this tab open" energy.
- Split inbox UI (sent bubbles, sidebar, smooth scrolling)
- Filter by All, DMs, or Groups, plus in-app search
- Contact names when available; push names when that is all WhatsApp provides
- Group chats show sender name and avatar per message
- Unread badges on new messages; clear when you open a chat
- Dark, Light, or system theme
- Local SQLite storage for sync and external access
- Relink via QR when your session or phone changes
| Layer | Technology |
|---|---|
| Shell | Electron 33 (frameless on Linux) |
| UI framework | React 19 + Vite |
| Component library | chatcn (macOS Messages style) |
| Styling | Tailwind CSS v3 |
| WhatsApp protocol | Baileys 6.x (main process only) |
| Database | better-sqlite3 (SQLite WAL) |
| IPC | contextBridge + typed channels (renderer never touches Node directly) |
Download from Releases:
| Platform | File |
|---|---|
| Linux (x86_64) | WhatsZapp-*-linux-x64.AppImage |
| Windows (x64) | WhatsZapp-*-windows-setup.exe |
Linux (AppImage):
chmod +x WhatsZapp-*-linux-x64.AppImage
./WhatsZapp-*-linux-x64.AppImageIf you see a FUSE error, install fuse2 (sudo pacman -S fuse2 on Arch, sudo apt install libfuse2 on Ubuntu/Debian).
Windows (installer):
Run the .exe and follow the prompts. SmartScreen may flag the unsigned binary on first run. Use More info → Run anyway if you trust the build.
Requirements: Node.js 20+, Linux or Windows (macOS works for dev; AppImage targets Linux)
git clone https://github.com/laxman-patel/WhatsZapp.git
cd WhatsZapp
npm install # postinstall rebuilds better-sqlite3 for ElectronDevelopment:
npm run devOpens Vite + Electron with hot reload. On first launch, link your account: WhatsApp → Settings → Linked Devices → Link a device, then scan the QR code.
Production build:
npm run build # output in release/<version>/Produces an AppImage on Linux and an NSIS installer on Windows.
Tagged releases (CI):
git tag v0.2.0
git push origin v0.2.0GitHub Actions uploads artifacts to the Releases tab.
Everything stays on your computer:
| Path | Contents |
|---|---|
~/.config/whatsapp-desktop/baileys-auth/ |
WhatsApp session keys (Linux) |
~/.config/whatsapp-desktop/whatsapp.db |
Messages, chats, contacts (SQLite) |
~/.config/whatsapp-desktop/avatars/ |
Cached profile pictures |
On Windows, the same folder name lives under %APPDATA%\whatsapp-desktop\.
The app talks to WhatsApp's servers for messaging only. This project does not run a cloud backend that sees your chats. When you add AI, you choose where that processing happens.
electron/
main/
baileys/ WhatsApp socket, avatar fetcher, contact aliases
db/ SQLite schema, migrations, repositories
index.ts Electron app entry, protocol registration
ipc.ts IPC handler registration
sync-progress.ts Sync state machine + broadcaster
src/
components/ React UI components
ui/chat/ chatcn message/sidebar/header components
hooks/ useChats, useMessages, useSyncProgress
lib/adapters/ Baileys → chatcn data mappers
shared/ IPC types shared between main & renderer
AI and automation
- Bring-your-own AI (local models or your API keys, not Meta's)
- Agent API (local HTTP / MCP endpoint for inbox read/write)
- Smart inbox (priority contacts, custom filters, rules)
- Programmatic send from scripts
- Webhooks on new messages
- Reply from the terminal
Messaging
- Media download and display (images, video, audio, documents)
- Send attachments from the composer
- OS notifications for new messages
- Read receipts via Baileys
- System tray icon and badge
- Full-text search across SQLite
- Per-chat export (JSON/CSV)
WhatsZapp is not affiliated with Meta or WhatsApp. It uses the unofficial Baileys library, which reverse-engineers the WhatsApp Web protocol. Unofficial clients may violate WhatsApp's Terms of Service and can lead to account restrictions or bans. Use at your own risk.
MIT. See LICENSE.
Application code © 2025 Laxman. Baileys is separately licensed under its own terms.