NGARi FAQ

NGARi is a sovereign AI platform that runs on private servers — no third-party AI providers process your data. No OpenAI, no Google, no external AI APIs. You own the model outputs, and your conversations are never sent to a third-party AI company. Self-hosted NGARi+ is available for those who want it on their own hardware.

Those services run on someone else's servers, and your conversations are stored and processed by their teams. NGARi runs on our private servers — your conversations are never sent to OpenAI, Google, or any third-party AI provider. There are no refusal filters and no API rate limits from a third party. For full sovereignty, NGARi+ runs entirely on your own hardware.

All you need is a web browser. NGARi Cloud runs on our servers — no installation, no hardware requirements. If you want NGARi on your own hardware, NGARi+ runs on NVIDIA Jetson Orin Nano (8GB+), Mac with Apple Silicon, or any Linux server with a GPU (8GB+ VRAM for 14B models, 24GB+ for 32B models).

All inference happens on our private servers — not OpenAI, not Google, not any third-party AI. Conversation data is stored in our PostgreSQL database on our private infrastructure. NGARi does not send your conversations to any third-party AI service. The only external network calls are for Stripe billing (if you subscribe) and magic link authentication emails — both of which are standard web infrastructure. For NGARi+ self-hosted deployments, these can be configured locally.

The free tier includes seven models: NGARi 1.5B (our sovereign model, the default), Qwen 2.5 0.5B, Llama 3.2 1B, Qwen 2.5 1.5B, Qwen 2.5 7B, GLM-4 9B, and Qwen 2.5 Coder 14B. The X plan adds Qwen 2.5 32B. The default is tuned for fast, accurate responses; switch to a larger model anytime from the model selector. All models run privately through Ollama — no third-party AI API keys. With NGARi+, you can add your own models too.

NGARi includes six specialized AI agents — Sage (Strategy), Maya (Revenue), Kai (Security), Imani (Operations), Zara (Marketing), and Ayo (Coordination). Each has its own expertise and persona. Turn on Power mode to pick an executive; the agent introduces itself and stays in its lane, handing off to another when a question fits them better. Use NGARi directly in Simple mode for everything else.

The free tier includes 50 conversations per month (10 for anonymous users), all seven free models, and all six agents. No credit card required. Upgrade to Build ($79/mo), Scale ($249/mo), or Company ($799/mo) for the 32B model, extended context windows, and more capacity.

NGARi Pay is sovereign, multi-rail payment infrastructure — one checkout for cards, Bitcoin, Lightning, agent (x402) payments, and ACH, running on your own hardware. Bitcoin settles self-custodially to a wallet you control, and every settlement produces a cryptographic, verifiable receipt. No processor can freeze you, and the keys are yours. Learn more at ngari.com/pay.

Yes — NGARi+ is a sovereign AI appliance you own outright, with the full stack pre-loaded on NVIDIA Jetson. Three tiers are available to pre-order now: NGARi+ Nano ($499, Jetson Orin Nano), NGARi+ Orin ($2,999, Jetson AGX Orin, runs the 32B model), and NGARi+ Thor ($5,999, Jetson AGX Thor flagship). Fully air-gapped, zero cloud, zero telemetry. Pre-order at ngari.com/plus.

NGARi is governed by construction, not by marketing promises. Our Tech Charter, Sovereign Constitution, and SBOS Manifesto define binding standards — zero cloud dependency, zero third-party inference, zero undisclosed telemetry, and human-in-the-loop authority. Read or download all three at ngari.com/governance.

NGARi uses magic-link authentication. Enter your email and we send a one-time sign-in link — no passwords to remember or leak. The email comes from our private mail server, not a third-party marketing service.

Sign in and click Manage in the sidebar to open your Stripe billing portal. Upgrade, downgrade, or cancel anytime; changes take effect at the end of your billing period. No lock-in, no cancellation fees.

NGARi Cloud needs connectivity to our servers for inference and conversation storage. Self-hosted NGARi+ runs fully offline — only billing and initial setup require internet access.