> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/aerial/aodt/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/aerial/aodt/_mcp/server.

# AODT Installation

This guide helps you install AODT on supported hardware and choose the right deployment model for your environment.

## Components

AODT consists of three components:

| Component  | Role                                          | Requirements                                              |
| ---------- | --------------------------------------------- | --------------------------------------------------------- |
| **Client** | Configures simulations and controls execution | No GPU required. Linux, Windows, and macOS are supported. |
| **Worker** | Runs the simulation                           | GPU-enabled Linux system required.                        |
| **Viewer** | Visualizes scenarios and results              | Modern web browser only.                                  |

## Deployment types

Installations are characterized by where the client and worker run on the network:

**Colocated** — The client and worker run on the same system. Supported only on Linux OS.

**Non-colocated** — The client and worker run on different systems. They may be on the same network or on separate networks.

## Worker system requirements

| Requirement | Supported values                       |
| ----------- | -------------------------------------- |
| **OS**      | Ubuntu Linux 22.04                     |
| **GPU**     | RTX 6000 Ada, L40, A100, H100          |
| **Driver**  | NVIDIA driver 575.57.0 (576.80 on AWS) |

## Qualified Worker systems

The following systems have been tested and are supported with the AODT installer.

### AWS VM (colocated)

* Instance type: `g6e.xlarge`
* OS: Ubuntu Server 22.04
* GPU: NVIDIA L40S
* vCPUs: 4
* Memory: 32 GB

### Dell R750 (colocated)

* OS: Ubuntu Server 22.04
* GPU: NVIDIA L40S or NVIDIA RTX 6000 Ada
* CPU: Intel Xeon Gold 6336Y 2.4 GHz, 24C/48T
* Memory: 512 GB DDR4
* Storage: 2 TB

## AWS installation

Set up an AWS VM using the configuration above as a reference. SSH access to the worker host is required before you begin installation. Once SSH access is available, the installation is identical to a colocated Linux installation.

The steps below walk through installing all three components.

## Operational modes

AODT supports two operational modes:

1. **Runtime mode** — Use prebuilt Docker images and precompiled binaries to run a full AODT instance.
2. **Development mode** — Modify source code and rebuild Docker images for client and worker binaries as needed.

The client and worker do not need to use the same mode. For example, you can run a prebuilt worker while developing the client locally.

## Installation overview

Follow these steps in order:

1. Install [prerequisites](/prerequisites) (GPU drivers, NVIDIA Container Toolkit, and related dependencies).
2. Download and unzip the AODT package from NGC.
3. Start AODT using prebuilt images and binaries, or optionally build the client and worker from source.
4. Verify the installation.
5. Install the viewer.

Install each component using the links below. Start with the client, then the worker, then the viewer.

Build and install the C++ and Python client library.

Start the Digital Twin server and infrastructure with Docker Compose.

Verify installation.

Set up the web viewer for simulation visualization.

After the components are installed, return to this page and choose a workflow for runtime or development use.

Run AODT with prebuilt Docker images and binaries.

Modify source code and rebuild client or worker components.