# How to Run OpenClaw with Linode

> Deploy OpenClaw on a Linode VPS and run AI agent with full privacy and control.

**Published:** Mar 20, 2026 | **Author:** Alex Chen | **Read time:** 12 min read

Run OpenClaw on a Linode VPS to set up a 24/7 AI agent with full control and privacy. Step-by-step from instance creation to systemd and firewall.

---

- OpenClaw with Linode' subtitle="Deploy OpenClaw on a Linode VPS and run AI agent with full privacy and control." readTime="12 min read" publishDate="Mar 20, 2026" author="Alex Chen" /> [OpenClaw](/blog/what-is-openclaw) is a self-hosted AI agent framework that runs on any Linux server. Deploying it on Linode gives you a private, always-on agent with full control over your API keys, conversation history, and configuration. If you prefer a managed setup, [Ampere.sh](https://www.ampere.sh) gets OpenClaw running in 60 seconds — no server required. Requirements Before starting, make sure your Linode meets these specs. An undersized plan is the most common reason installs fail. Component Requirement **Linode Plan**Linode 4 GB (2 vCPU / 4 GB RAM) minimum — 8 GB for multi-agent **OS Image**Ubuntu 24.04 LTS or Ubuntu 22.04 LTS **Node.js**v22.16 or higher — install via NodeSource, not apt **SSH Access**SSH key authentication — disable password auth after first login **Firewall**UFW — allow SSH (22), deny port 18789 from public access **Storage**25 GB disk minimum — default Linode disk is enough Step 1: Create Your Linode Log in to the [Linode Cloud Manager](https://cloud.linode.com) and click **Create → Linode**: **Image:** Ubuntu 24.04 LTS x64

[OpenClaw](/blog/what-is-openclaw) is a self-hosted AI agent framework that runs on any Linux server. Deploying it on Linode gives you a private, always-on agent with full control over your API keys, conversation history, and configuration. If you prefer a managed setup, [Ampere.sh](https://www.ampere.sh) gets OpenClaw running in 60 seconds — no server required.

[OpenClaw](/blog/what-is-openclaw) is a self-hosted AI agent framework that runs on any Linux server. Deploying it on Linode gives you a private, always-on agent with full control over your API keys, conversation history, and configuration. If you prefer a managed setup, [Ampere.sh](https://www.ampere.sh) gets OpenClaw running in 60 seconds — no server required.


## Requirements


| Component | Requirement |
| --- | --- |
| Linode Plan | Linode 4 GB (2 vCPU / 4 GB RAM) minimum — 8 GB for multi-agent |
| OS Image | Ubuntu 24.04 LTS or Ubuntu 22.04 LTS |
| Node.js | v22.16 or higher — install via NodeSource, not apt |
| SSH Access | SSH key authentication — disable password auth after first login |
| Firewall | UFW — allow SSH (22), deny port 18789 from public access |
| Storage | 25 GB disk minimum — default Linode disk is enough |

Before starting, make sure your Linode meets these specs. An undersized plan is the most common reason installs fail.


## Step 1: Create Your Linode

Log in to the [Linode Cloud Manager](https://cloud.linode.com) and click **Create → Linode**:

- **Plan:** Linode 4 GB for personal use — Linode 8 GB for multi-agent
- **Region:** Pick the closest one for low latency
- **Authentication:** SSH Key — add your public key during creation
- **Label:** `openclaw-agent`

Once the Linode boots, SSH in and create a non-root user:


## Step 2: Install Node.js

OpenClaw requires Node.js v22.16 or higher. Never use `apt install nodejs` — it installs an outdated version. Always install from NodeSource:

Running on Windows? [Follow the WSL guide](/blog/how-to-run-openclaw-on-wsl) to run OpenClaw on Windows Subsystem for Linux instead.


## Step 3: Install OpenClaw

Run the official installer — it detects Node and installs the CLI globally:

Prefer to inspect the script first:

If `openclaw` is not found after install, fix PATH:

Once installed, explore [the best OpenClaw skills](/blog/best-openclaw-skills) or learn how to [build a custom skill](/blog/openclaw-custom-skill).


## Step 4: Onboarding & Connect Your Channels


| Channel | Setup Guide |
| --- | --- |
| Telegram | Connect OpenClaw to Telegram → |
| WhatsApp | Connect OpenClaw to WhatsApp → |
| Discord | Connect OpenClaw to Discord → |
| iMessage | Connect OpenClaw to iMessage → |
| Notion | Connect OpenClaw to Notion → |

Run the onboarding wizard — it configures your AI provider, connects a channel, and installs the systemd daemon:

The wizard walks you through selecting your AI provider, entering an API key, choosing a model, and connecting a channel. OpenClaw supports all major channels:

Verify your agent is live:


## Step 5: Enable systemd & Configure Firewall

Enable linger so the service starts on boot without an active SSH session, then start the gateway:

Now lock down the firewall. Port 18789 must never be publicly accessible. For a full hardening checklist, see the OpenClaw security checklist and the complete security guide.

To update OpenClaw:


## Troubleshooting

Most issues come down to four things: Node.js version, PATH, memory, and systemd linger.


### `openclaw: command not found`

npm installed OpenClaw globally but the global bin directory is not in your PATH. Fix it permanently:


### npm install killed — exit code 137 (OOM)

You ran out of memory during install. This happens on the 1 GB or 2 GB Linode. Add swap before retrying:


### `node --version` shows v12 or v18 (too old)

You installed Node.js from apt instead of NodeSource. Remove and reinstall:


### Gateway does not start on reboot

Linger is not enabled — user services only start when you are logged in without it:


### systemd service fails — ExecStart path not found

The unit file references the wrong binary path. Find the correct path and update:


### Port 18789 refused or gateway unreachable

The gateway binds to 127.0.0.1 by default — it is not reachable publicly on purpose. Access it via SSH tunnel:


## Linode vs Ampere.sh


| Feature | Linode (Self-Hosted) | Ampere.sh (Managed) |
| --- | --- | --- |
| Setup Time | 1–2 hours | 60 seconds |
| Technical Skill | Linux, SSH, systemd | None required |
| Server Management | You manage everything | Fully managed |
| Uptime | Depends on VPS health | 99.9% SLA |
| Maintenance | Manual updates | Automatic |
| Data Privacy | Your server, full control | Managed cloud |
| Cost | Server cost + your time | Free credits to start |

Running OpenClaw on Linode gives you full control but requires server management. Ampere.sh is the managed alternative — same OpenClaw, zero infrastructure. Also worth reading: AI agent hosting options compared.


## Skip the Server Setup?

Deploy OpenClaw in 60 seconds on Ampere.sh — no VPS, no systemd, no firewall config. Free credits to get started.

[Deploy on Ampere.sh →](https://www.ampere.sh/setup)


## Frequently Asked Questions

### Which Linode plan should I use for OpenClaw?

The Linode 4 GB (2 vCPU / 4 GB RAM) is the recommended starting point for personal use. The 2 GB plan can work with swap but leaves little headroom. Use the 8 GB plan for multiple agents or heavier models.

### Does OpenClaw work on Ubuntu 22.04 LTS on Linode?

Yes. Both Ubuntu 22.04 LTS and Ubuntu 24.04 LTS are fully supported. The commands in this guide apply to both.

### Will OpenClaw keep running after I close my SSH session?

Yes — once you enable linger and the systemd service in Step 5, OpenClaw runs as a background daemon. It starts on boot and restarts on crash without an active SSH session.

### Can I run OpenClaw on a Linode Nanode (1 GB)?

Not reliably. The npm install process requires more than 1 GB RAM and will be killed by the OOM killer. Add a 2 GB swap file to get through the install, but upgrade to the 4 GB plan for stable use.

### Do I need a domain for OpenClaw on Linode?

No. OpenClaw connects to messaging platforms outbound — no inbound port exposure needed. A domain is only required if you want a custom HTTPS nginx endpoint for the gateway UI.

### How do I update OpenClaw on Linode?

Run npm install -g openclaw@latest, then restart: systemctl --user restart openclaw-gateway. Keep system packages updated too with sudo apt update && sudo apt upgrade -y.

### Can I run multiple agents on one Linode?

Yes, on a Linode 8 GB or higher. Each agent is a separate Node.js process with its own systemd unit file and port assignment.

### Is the installer script safe to run?

Yes. To review it first: curl -fsSL https://openclaw.ai/install.sh -o install.sh, inspect with less install.sh, then bash install.sh. Alternatively: npm install -g openclaw@latest.