IK
09AI TOOLS2026IN DEVELOPMENT

SANAD AI OPERATOR

سندك الذكي… يسمع ويفهم وينفّذ.

SANAD AI logo

A local, Arabic-first, voice-controlled desktop AI operator: wake word «يا سند», browser and desktop automation, email, messaging, workflows, RAG memory, approval-gated actions, an admin console and Stripe billing — FastAPI backend, React + Tauri desktop.

SANAD AI Operator reached v4.0 'Stable Operator' after fourteen phases: RTL foundation with approvals and WebSocket; Arabic voice over OpenAI Realtime with a wake word; Playwright browser automation; Gmail and Outlook OAuth; desktop control with allowlisted commands and pywinauto; memory and a sequential workflow engine; WhatsApp and Telegram web composition; an APScheduler autopilot with a morning briefing; Tesseract OCR vision; a dev operator for git and editors; enterprise policy rules and audit export. Later phases added a ChatGPT-like universal chat with multiple providers and seven answer modes, live web search with SSE streaming, TTS readback, export, a Tauri desktop bundle, an enterprise dashboard and a plugin registry with webhooks.

The backend is Python/FastAPI with SQLAlchemy, Playwright, pyautogui, APScheduler, pytesseract and pywebpush; the desktop is React 19 + Vite 6 + Tauri 2 with 25 pages. CI runs on GitHub Actions and the release checklist, Stripe live setup and Tauri build are documented.

01

Overview

An operator, not a chatbot. Sanad listens in Arabic, plans, asks for approval where policy requires it, and executes across the browser, the desktop, mail and messaging.

SANAD AI logo
02

Architecture

  1. VOICE
    OpenAI Realtime · wake word
  2. PLANNER + ROUTER
    agent loop · tool registry
  3. POLICY GATE
    approvals · trusted sessions
  4. TOOLS
    browser · desktop · email · messaging · dev
  5. MEMORY / RAG
    semantic memory · Excel RAG
  6. AUTOPILOT
    APScheduler · morning briefing
  7. BILLING · ADMIN · ANALYTICS
    Stripe · enterprise
03

Capabilities

  • 01Universal AI chat — multi-provider, seven answer modes, SSE streaming, image input
  • 02Browser automation — multi-tab navigate, search and screenshot with Playwright
  • 03Desktop control — open apps, screenshots, allowlisted commands
  • 04Email — Gmail and Outlook OAuth read/send; messaging via WhatsApp Web and Telegram
  • 05Workflows and replay; long-term memory («تذكر أن…»)
  • 06Security Center, custom policy rules, audit export
  • 07Tauri installer for Windows; portable bundled backend
04

Brand

Next project
NEXIO ERP
Rendered vertical video with synced Arabic captions
08PROTOTYPE
AI TOOLS · 2026

LYRICCAPTION

HTML / CSS / JAVASCRIPTI18N / RTL ENGINEERINGFASTAPIPYTHON

Speech to synced Arabic captions, locally.