OpenJet – Agentic Terminal UI for local LLMs on NVIDIA Jetson devices

Summary: OpenJet is a terminal user interface designed for self-hosted large language models on NVIDIA Jetson devices, enabling fully local operation without internet or cloud dependencies. It supports unified memory systems to prevent out-of-memory errors and keeps all inference and logs on-device for edge computing.

What it does

OpenJet runs AI agents locally on Jetson hardware, allowing file editing, reading, and data management without cloud access. It achieves about 17 tokens per second on Jetson Orin Nano 8GB using Qwen3-4B-Instruct-4bit and plans to add TensorRT engine support for faster inference.

Who it's for

It is intended for users deploying self-hosted AI agents in air-gapped or edge environments using NVIDIA Jetson devices with unified memory architectures.

Why it matters

OpenJet solves latency, privacy, and connectivity issues by enabling fully local LLM inference and data handling on edge devices without cloud reliance.