SELF-HOSTED
NGARi+ — AI on Your Hardware
NGARi+ runs the entire stack on your own infrastructure. No cloud. No third-party AI providers. Fully air-gapped. Your hardware, your network, your rules.
Fully Air-Gapped
Run without internet. No cloud calls, no phone-home, no telemetry. Only you have access to your data.
Zero Cloud Dependencies
Next.js, PostgreSQL, Ollama — all running locally in Docker containers you can inspect, audit, and modify.
Same AI, Your Rules
All six NGARi agents, all models (including Qwen 2.5 32B), the full conversation system — on your terms.
Recommended Hardware
| Hardware | GPU Memory | Supported Models |
|---|---|---|
| NVIDIA Jetson Orin Nano | 8 GB | 7B, 9B, 14B (quantized) |
| Mac Mini M2 Pro+ | 16 GB unified | All models through 14B |
| Linux Server (RTX 3090+) | 24 GB | All models including 32B |
| Cloud VM (GPU instance) | Varies | Depends on GPU |
Join the NGARi+ Waitlist
NGARi+ is coming soon. Enter your email to be notified when self-hosted deployment is available.
NGARi Cloud vs NGARi+
| NGARi Cloud | NGARi+ | |
|---|---|---|
| Private inference | On our servers | On your hardware |
| Third-party AI | None | None |
| Internet required | Yes | No (after setup) |
| Data sovereignty | We host your data | You host your data |
| Setup effort | None (browser) | Docker deploy |
| Models | Same | Same + custom |
| Price | Free / $79 / $249 / $799 | TBD |
Not ready for self-hosting? Try NGARi Cloud free.
Try NGARi Cloud