---
title: "Set Up RTX Spark as Your Always-On AI Hub — Agents"
canonical: "https://build.nvidia.com/playbooks/ai-hub/agents.md"
---

# Run your agent on RTX Spark

Your agent and its model both run on RTX Spark, so the agent stays available when your other devices are off. You talk to it from your desk, your phone, or another PC. If you want to keep the agent on your laptop and use RTX Spark only for the model, use the **Inference** tab instead.

**Before you start:** complete the **Remote Access** tab.

# Step 1. Install your agent on RTX Spark

On RTX Spark, follow these steps of [Set Up Your First AI Agent](https://build.nvidia.com/rtx-spark/start-with-agent#:~:text=Choose%20%26%20install%20your%20agent), then return here:

- **Step 1. Choose & install your agent.**
- **Step 2. Choose your model.** Pick a model under **LOCAL** so that it runs on RTX Spark.
- **Step 4. Talk to your agent.** Send one message to confirm that the agent answers.

Steps 3 and 5 of that playbook are optional. You can come back to them later.

# Step 2. Keep your agent running

Your agent answers only while its app is running on RTX Spark.

- **Leave the agent app open.** When you finish a remote desktop session, close the remote desktop window. Do not sign out of Windows, because that closes the agent.
- **Start the agent when you sign in.** Turn on your agent app in **Settings > Apps > Startup** if it is listed there, or use the agent’s own setting to start at sign-in.
- **After a restart, sign in once.** Apps start only after someone signs in to Windows. Connect with remote desktop, sign in, and check that the agent app is open.

# Step 3. Connect from your other devices

- **From your phone.** Connect a messaging app, such as Telegram or Discord, to your agent. Follow [Channels](https://build.nvidia.com/rtx-spark/start-with-agent#:~:text=Channels%20%2D%20Connect%20to%20your%20agent%20from%20anywhere) in the agent playbook, then return here. Some agents also have a mobile app, such as [OpenClaw for iOS](https://docs.openclaw.ai/platforms/ios).
- **From another PC.** Connect with remote desktop (**Remote Access** tab, Step 3) and use the agent app on RTX Spark’s desktop.
- **At your desk.** Connect a monitor, keyboard, and mouse to RTX Spark and use it like any other PC.

## Use the agent app on another PC (advanced)

Some agent apps can run on another PC as a window into the agent on RTX Spark. The agent, its tools, and its memory stay on RTX Spark. Each agent calls the part that runs on RTX Spark its gateway or backend.

Setup differs for each agent and changes between releases, so follow the agent’s own guide:

- Hermes: [Connecting to a remote backend](https://hermes-agent.nousresearch.com/docs/user-guide/desktop#connecting-to-a-remote-backend)
- OpenClaw: [Remote access](https://docs.openclaw.ai/gateway/remote)

Two details are specific to RTX Spark:

- **Address.** Where the guide asks for the remote address, use RTX Spark’s device name on your home network. To connect from anywhere, use its Tailscale name (**Remote Access** tab, Step 4).
- **Restarts.** Other PCs can connect only while the gateway is running on RTX Spark. After a restart, check that it is running before you connect.

> [!NOTE]
> This connection can run commands on RTX Spark and read your agent’s API keys. Keep it on your home network or behind Tailscale, and never expose it directly to the internet.

# Step 4. Check your setup

1. From your phone or another PC, send your agent a message. It should answer.
2. Restart RTX Spark, sign in through remote desktop, and send another message. The agent should answer again.