Your computer, as a conversation.
An open-source, AI-native operating system. Voice-first. Local-first. Yours.
What is this? • How it feels • Architecture • Quick Start • Screenshots • Roadmap • Contributing
OS 1 is a different kind of operating system. There are no desktop icons. No start menus. No app launchers. When you turn on your computer, there is one thing: a warm, conversational presence that listens, understands, and helps.
You talk to your computer. It talks back. It remembers what you were working on. It notices patterns. It generates whatever interface you need, right when you need it.
Traditional OS: You click icons → open apps → navigate menus → do the thing
OS 1: You say what you need → it appears → done
OS 1 is built entirely from open-source components. It runs locally on your machine — your voice, your data, your conversations never leave your device. It's written in Rust. It's built on Linux. And it's completely hackable.
Think of it as a platform. Drop in your own models, swap voice engines, build custom tools, reshape the experience. OS 1 is the foundation — what you build on it is yours.
OpenClaw showed what happens when a personal AI agent is open-source, hackable, and runs on your own machine — people made it theirs. OS 1 is that moment, one level down: not an agent running on your OS, but the OS itself.
OS 1 is inspired by the warmth and intimacy of conversational AI — a computer that feels less like a tool and more like a companion. This is the experience we're building toward:
- Morning. You boot your machine. OS 1 greets you. "Good morning. You have a meeting at 10. That PR from last night got merged."
- Working. "I need to fix the auth bug." OS 1 finds the repo, opens the files, dispatches an agent, shows you the diff. One tap to commit.
- Stuck. "My disk is filling up." A treemap appears. Caches highlighted. "Clean" buttons next to each one.
- Curious. "What was I working on last Tuesday?" OS 1 remembers. It shows you.
- Done. "Play something chill." Minimal controls materialize. No app. Just music.
There are no applications in the traditional sense. OS 1 generates fragments — purpose-built interface components that appear when needed and fade when done.
Where we are honestly: v1 boots today — a conversational shell with voice in/out, 12 sandboxed system tools, agent dispatch, and a privacy filter, all running on local models. Generated fragments, persistent memory, and the proactive companion are v2, being built in the open. Boot the ISO and you get a real conversational OS, not the full movie — yet.
 Boot screen |
Generative Desktop |
 Installing to disk |
|
┌──────────────────────────────────────────────────────────────────┐
│ GENERATIVE SHELL │
│ Iced + Wayland layer-shell — voice input, fragment rendering │
├──────────────────────────────────────────────────────────────────┤
│ VOICE I/O │
│ Piper (TTS) │ Faster-Whisper (STT) │ OmniVoice-Studio (v2) │
├──────────────────────────────────────────────────────────────────┤
│ ORCHESTRATOR — The Brain │
│ Intent → Model Selection → Agent Dispatch → Tool Execution │
│ Semantic memory across sessions (LanceDB → fast-mempalace v2) │
├────────────┬────────────┬─────────────┬──────────────────────────┤
│ LOCAL │ AGENT │ MCP TOOLS │ PRIVACY FILTER │
│ MODELS │ HARNESS │ 12 system │ PII never leaves │
│ Ollama │ OpenCode │ tools with │ your device │
│ Any model │ Claude │ sandboxing │ │
├────────────┴────────────┴─────────────┴──────────────────────────┤
│ MEMORY │
│ LanceDB + local embeddings — persistent, semantic, 100% local │
│ (fast-mempalace integration coming in v2) │
├──────────────────────────────────────────────────────────────────┤
│ SECURITY │
│ Capability sandbox │ Audit log │ Path/command filtering │
├──────────────────────────────────────────────────────────────────┤
│ DESKTOP │
│ Niri (Wayland compositor) │ Waybar │ PipeWire │ NetworkManager │
├──────────────────────────────────────────────────────────────────┤
│ LINUX — CachyOS kernel / BORE scheduler │
└──────────────────────────────────────────────────────────────────┘
| Traditional OS | AI-Assisted OS | OS 1 | |
|---|---|---|---|
| Interface | Windows, icons, menus | Same + chat sidebar | Voice → generated UI |
| AI role | None | Copilot / assistant | The AI is the shell |
| Voice | "Hey Siri" gimmick | Basic commands | Primary interaction mode |
| Memory | None | Cloud-dependent | Local semantic memory |
| Data | Cloud sync | Sent to APIs | Never leaves your machine |
| Hackability | Closed | Plugin APIs | Fork it, reshape it, own it |
OS 1 ships with a curated stack, but everything is swappable:
| Layer | Ships today | Coming / swap in anything |
|---|---|---|
| Voice | Piper (TTS) + Faster-Whisper (STT) | OmniVoice-Studio (v2), Bark, XTTS |
| Memory | LanceDB + nomic-embed-text | fast-mempalace (v2), ChromaDB, Qdrant |
| Models | Ollama (qwen, llama, gemma) | Any GGUF, any API |
| Agents | OpenCode, Claude Code | Codex CLI, custom agents |
| Compositor | Niri | Hyprland, Sway, any Wayland WM |
| Browser | Donut Browser | Firefox, Chromium |
- Download the latest ISO from Releases
- Flash to USB:
sudo dd if=os1-*.iso of=/dev/sdX bs=4M status=progress - Boot from USB
- Start talking
git clone https://github.com/debpalash/OS1.git
cd OS1
# Build everything in the Cargo workspace
cargo build --release --workspace
# ...or a single crate
cargo build --release -p indos-orchestrator
# Build bootable ISO (needs archiso + root)
sudo ./indos-iso/build-iso.sh
# Run automated tests
python3 autotest/autotest.py| Category | Technology |
|---|---|
| Language | Rust (core), Python (voice), TypeScript (fragments) |
| Base | Arch Linux (CachyOS kernel, BORE scheduler) |
| Compositor | Niri — scrollable tiling Wayland |
| Shell | Iced 0.14 + Wayland layer-shell |
| Voice | Piper (TTS) + Faster-Whisper (STT) — OmniVoice-Studio in v2 |
| Memory | LanceDB + nomic-embed-text — fast-mempalace in v2 |
| Inference | Ollama (fully offline) |
| Audio | PipeWire |
| Protocol | MCP / NDJSON over Unix socket |
- Orchestrator with intent classification and model routing
- Generative shell (Iced + Wayland layer-shell)
- 12 MCP tools with security sandbox
- Voice pipeline (STT + TTS)
- Privacy filter with PII redaction
- Bootable Live ISO with installer
- Automated test suite (22 stages)
- CI/CD for ISO builds
- HER-inspired warm UI theme (coral/amber palette)
- Plymouth animated boot splash — breathing coral ring, HER-style
- Voice-first interaction mode —
indos-voiced: hands-free wake-phrase loop ("hey OS"), conversation window, push-to-talk (Mod+Shift+V) - Proactive daily briefing — spoken greeting with system facts at first login of the day
- OmniVoice-Studio integration — custom voice personalities
- fast-mempalace integration — persistent semantic memory
- Pattern recognition — ambient context for proactive help
- Fragment marketplace — community-built UI components
- Custom voice packs
- Multi-agent orchestration
- ARM64 support
- Self-healing system diagnostics
- Multilingual voice support
OS 1 is a hackable platform. We welcome contributions of all kinds — code, voice packs, fragments, translations, and ideas.
See CONTRIBUTING.md for setup and guidelines.
git clone https://github.com/YOUR_USERNAME/OS1.git
cd OS1
git checkout -b feature/your-feature
# Make it yours
git push origin feature/your-featureGPL-3.0 — See LICENSE
An open-source project. Built with ❤️ in India.
Star ⭐ if you believe the desktop deserves to feel human.