The AI-native personal OS you can run today. Pure Python neuros — no new language required. Voice in, voice out. Web, desktop, and Android. Multi-agent rooms, scheduler, 3D IDE, terminal.
Neos runs the same Python backend across all surfaces. Switch between them seamlessly.
Next.js + R3F frontend. Classic tabs or 3D spatial view. Split panes. Runs at localhost:3000. PWA-installable.
Electron wrapper. Native window controls. Launches backend automatically. Same UI as web, no browser required.
Jetpack Compose client. Metro-style tile launcher. Voice typing. Remote desktop via WebRTC. Pinch zoom up to 10×.
Everything wired together out of the box.
STT/TTS round-trip with Sarvam streaming + ElevenLabs / OpenAI synthesis. Runs as a system service. Agents speak, listen, and converse with sub-second latency.
Author neuros visually with @react-three/fiber. Source-aware editing of conf.json, code.py, prompt.txt. Hot-load on save.
Multiple agents share a transcript with a mediator picking the next speaker round-robin. Live in room_* neuros.
Cron-style triggers persisted via APScheduler. Set timers, recurring digests, deferred tasks — all from the chat surface.
Direct typed messages between agents with depth-guard (MAX=4). The substrate beneath rooms.
Swap model, system prompt, defaults at the agent level. Ships with general, code_dev, neuro_dev, neurolang_dev.
Each app is a neuro-backed agent with a focused persona. Pinned to the dock or discoverable from the launcher.
General AI assistant
Web automation
Code assistant
Upwork automation
Deep research
Writing assistant
Data analysis
File manager
Email assistant
Scheduling AI
Voice processing
AI translation
Neos runs on a Python backend (port 7000) + Next.js frontend (port 3000). Android client connects to the same backend.