sonichi/sutando
My AI Stand. Realtime by day, rewriting itself by night. Summon my AI superpower.
What it solves
Sutando solves the problem of having a personal AI that can act in real time—through voice, screen, and calls—while also working autonomously in the background to improve itself. It replaces the need to manually handle routine tasks like joining meetings, making phone calls, or editing documents by letting you talk to your computer and have it execute actions across your system. It also addresses the pain of maintaining an AI assistant by letting it continuously learn your patterns and write its own code to add missing capabilities.
How it works
Sutando runs on your Mac (with core headless support on Windows) using either Claude Code or Codex CLI as the execution engine. It consists of four main processes: a voice agent for real-time conversation (via Gemini Live), a web client for the browser UI, a conversation server for phone calls, and a core agent that executes tasks with full system access. A cron job fires a proactive loop every 15 minutes to process queued tasks, run health checks, and trigger autonomous improvements. Tasks are written to a tasks/ folder and results are written back, with each channel (voice, Telegram, Discord, etc.) reading from or writing to these files. It integrates with external services like Gemini for voice, Twilio for phone calls, and messaging bridges, all optional and independently enabling features.
Who it’s for
Sutando is for individuals who want a personal AI that can handle both real-time interactions (voice commands, meetings, calls) and background autonomous work like code writing and system monitoring. It suits power users comfortable with local AI setup, willing to use Claude Code or Codex CLI, and open to testing an alpha-stage product with 30 verified capabilities and a few requiring external setup (Twilio, Telegram bot, Discord bot). It’s also for developers interested in contributing to harden features and expand its capabilities.
Highlights
- Open source and runs entirely locally on your Mac; no per-token API key required.
- Voice and vision: sees your screen, responds to voice commands, can rewrite text and control apps.
- Joins meetings (Zoom, Google Meet) and makes phone calls, with a Telegram approval gate for task delegation.
- Autonomous build loop: when idle, it monitors its own health, discovers new skills, and writes its own code—most of Sutando’s code was written this way.
- Multi-channel access: same assistant via Telegram, Discord, web, phone, email, and more.
- Scales across machines: can set up a second Mac to handle calls 24/7 with no migration scripts.
- Early alpha status: 30 verified working features, but actively seeking contributors to test and harden capabilities.
Related
- Project
- Project
- Project
- Project
- Project