Skip to main content
agtOS is a platform, not a single-purpose app. Here are the primary ways people use it.

​
Personal AI Assistant

Use agtOS as your always-available voice assistant that actually remembers you.

Voice interaction

Talk naturally via the desktop app, browser dashboard, or ESP32 hardware. Press your push-to-talk hotkey from any application.

Persistent memory

agtOS remembers your preferences, past conversations, and learned facts across sessions. Ask “What did we discuss last week?” and get real answers.

Task scheduling

“Remind me every morning at 7 AM to check my calendar” — agtOS creates and manages cron-based, one-time, and interval tasks.

Privacy-first

Everything runs on your machine. Use Ollama for fully local AI. Cloud is optional and only used when you choose.
Example workflow:
  1. Install the desktop app and configure a Claude API key
  2. Say “Remember that I prefer metric units and morning meetings”
  3. Ask “Schedule a daily briefing at 7 AM”
  4. Next day, ask “What are my preferences?” — agtOS recalls everything

​
Smart Home Voice Hub

Connect agtOS to smart home devices via MCP for voice-controlled automation. 
You: "Turn off the kitchen lights and set the thermostat to 20 degrees"
     ↓
agtOS: Classifies intent → calls home.lights_off + home.set_temperature
     ↓
TTS: "Done. Kitchen lights are off and thermostat is set to 20°C."
How it works:
  • Connect an external smart home MCP server to agtOS
  • agtOS discovers available tools automatically
  • Voice commands are routed through the agent loop with tool execution
  • Responses are spoken back via TTS
Hardware: Use an ESP32-S3 as a dedicated voice terminal in each room. It streams audio over WebSocket to agtOS running on any machine in your network.

​
Developer Tool Integration

Connect agtOS to your development workflow as an MCP server for Claude Desktop, Cursor, or any MCP client.

Voice coding assistant

Talk through architecture decisions while agtOS records, remembers context, and integrates with your IDE.

Claude Desktop bridge

Add agtOS as an MCP server in Claude Desktop. Use voice, scheduling, and memory tools directly from Claude.
Claude Desktop integration: 
{
  "mcpServers": {
    "agtos": { "url": "http://localhost:4100/mcp" }
  }
}
Now Claude Desktop can:
  • voice.speak — read responses aloud
  • schedule.create — set reminders and recurring tasks
  • memory.ask_about_user — recall your preferences and past context

​
IoT Voice Device

Build a standalone voice device using the ESP32-S3 and agtOS as the backend brain. Hardware: Seeed Studio XIAO ESP32-S3 Sense with I2S microphone and DAC speaker. Architecture: 
ESP32 (mic + speaker)
    ↓ WebSocket (PCM audio)
agtOS Server (STT → LLM → TTS)
    ↓ WebSocket (PCM audio)
ESP32 (plays TTS response)
Features:
  • Wake word detection on-device
  • mDNS zero-config discovery (no manual IP configuration)
  • Device registry with per-device authentication and preferences
  • Platform-aware tool filtering (only send relevant tools to constrained devices)
See the firmware repository for ESP32 firmware code.

​
Multi-Agent Orchestration

Use agtOS as the voice and tool layer for multi-agent systems. The MCP server exposes 10 tools that any AI client can call, while the MCP client connects to external tool servers. 
Claude Desktop ──► agtOS (MCP Server) ──► Smart Home MCP Server
                       │
                       ├──► Calendar MCP Server
                       │
                       └──► Knowledge Base MCP Server
The agent reasoning loop handles multi-step tool execution automatically. Ask a complex question and agtOS orchestrates across multiple tool servers to build a response.