# Sutando

> An open-source, self-hosted AI agent for macOS that uses voice, vision, and autonomous action to control your computer, join meetings, make phone calls, and build itself.

Sutando is an open-source, self-hosted AI personal agent for macOS that combines voice control, screen vision, and autonomous task execution into a single local system. It runs on your existing Claude Code subscription and a free Gemini API key, with no remote control plane or third-party write access. Sutando can see your screen, join your meetings, make phone calls, send messages, and autonomously improve its own capabilities when idle — all from your Mac.

- **Voice control** — *Connect via browser or phone; say commands like "what's on my screen?" or "fix the typo in that file" and Sutando acts immediately using Gemini Live real-time voice.*
- **Screen vision** — *Sutando captures and analyzes your screen on demand, enabling context-aware assistance without manual copy-paste.*
- **Meeting participation** — *Say "join my 2pm call" and Sutando reads your calendar, joins Zoom or Google Meet with computer audio, researches questions live, and writes a summary when done.*
- **Phone calls** — *Sutando can make and receive calls via Twilio, have conversations on your behalf, and report back while you keep working.*
- **Multi-channel messaging** — *Reach the same agent via voice, Telegram, Discord, web, phone, or email — all sharing the same memory and task queue.*
- **Autonomous build loop** — *When idle, Sutando monitors its own health, detects usage patterns, discovers new skills, and writes missing capabilities — most of its own code was built this way.*
- **Notes and memory** — *Capture ideas by voice; Sutando tags, saves, and searches them as YAML-frontmatter markdown notes and acts on actionable items automatically.*
- **Multi-machine scaling** — *Plug in a second Mac and Sutando migrates services autonomously via Discord, coordinating the handoff between agents without migration scripts.*
- **3-tier access control** — *Owner, verified, and unverified callers get different capability bands on phone, Discord, and Telegram, with STIR/SHAKEN caller ID verification for inbound calls.*
- **Quick start** — *Clone the repo, add your `GEMINI_API_KEY` to `.env`, and run `bash src/startup.sh`; a menu bar app, dashboard, and voice interface launch automatically.*

## Features
- Voice control via browser or phone
- Screen capture and vision analysis
- Autonomous meeting joining (Zoom, Google Meet)
- Outbound and inbound phone calls via Twilio
- Multi-channel messaging (Telegram, Discord, email, web)
- Autonomous build loop and self-improvement
- Voice-driven note capture and search
- Multi-machine agent scaling and migration
- 3-tier access control (owner/verified/unverified)
- Global keyboard shortcuts via macOS menu bar app
- Proactive health monitoring and auto-repair
- Pattern detection and user modeling
- Gmail read/send/search
- Calendar and reminders integration
- Browser automation via MCP tools
- Cross-node memory and notes sync
- Info-radar (arXiv, GitHub, HN, news monitoring)
- System dashboard at localhost:7844

## Integrations
Claude Code, Gemini Live API, Twilio, ngrok, Telegram, Discord, Gmail (Google Workspace OAuth), Google Calendar, Zoom, Google Meet, macOS Contacts, macOS Reminders, Claude for Chrome (browser extension), WhatsApp (wacli)

## Platforms
MACOS, WEB, CLI, API

## Pricing
Open Source

## Version
v0.1-demo

## Links
- Website: https://sutando.ai
- Documentation: https://github.com/sonichi/sutando
- Repository: https://github.com/sonichi/sutando
- EveryDev.ai: https://www.everydev.ai/tools/sutando