Deploying OpenClaw via SSH Agent Commands

Deploying OpenClaw via SSH Agent Commands | Generated by AI

Home 2026.03

Question: What is the best way to use an existing OpenClaw AI agent to automatically install and set up another OpenClaw instance on a different machine, using only an SSH connection?

Answer:

Using OpenClaw Agent to Self-Deploy to Another Machine via SSH

Your idea is genuinely elegant — since OpenClaw already has terminal/shell access capabilities, you can simply instruct your existing agent to do the remote installation over SSH. This is exactly the kind of task it’s built for.

Prerequisites

Before telling your agent to go install itself remotely, make sure:

SSH connection from Machine A → Machine B is working (key-based auth preferred, no password prompt)
Machine B has internet access (to pull from npm)
Node.js >= 22 is required on the target machine (though the OpenClaw installer can handle this)
Minimum specs on the target: 2 cores, 2 GB RAM, 2 GB storage

What to Tell Your Existing OpenClaw Agent

Just chat with your agent naturally. Example prompt:

SSH into user@192.168.1.50 and install OpenClaw there.
Create a dedicated user called openclaw, install Node 24 via nvm,
then run the OpenClaw installer. Don't do the onboarding wizard yet.
I'll copy my config over after.

Your agent will then execute commands like:

# Step 1: Agent SSHes into target machine
ssh user@192.168.1.50

# Step 2: Create dedicated user (security best practice)
adduser openclaw --gecos "" && usermod -aG sudo openclaw

# Step 3: Install Node via nvm
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh | bash
source ~/.bashrc && nvm install 24

# Step 4: Install OpenClaw globally
npm install -g openclaw@latest

A dedicated non-root user is recommended — avoid running everything as root for better security.

After Installation: Copy Your Config (Fastest Path)

Once OpenClaw is installed on the target, have your agent copy your config over:

# Agent runs this on Machine A to copy config to Machine B
scp ~/.openclaw/openclaw.json user@192.168.1.50:~/.openclaw/openclaw.json

To fully migrate a setup including memory and sessions, copy $OPENCLAW_STATE_DIR (default: ~/.openclaw) and the workspace directory, then run openclaw doctor and restart the gateway service. This preserves config, auth profiles, channel credentials, sessions, and memory.

Then Have the Agent Start the Daemon

# Agent SSHes back in and runs:
ssh openclaw@192.168.1.50 "openclaw onboard --non-interactive --install-daemon --skip-skills"

Or if you just want the gateway started without full onboarding:

ssh openclaw@192.168.1.50 "openclaw gateway start"

Full Flow Summary

[Machine A - Existing Agent]
        |
        | SSH + shell commands
        ↓
[Machine B - Target]
  1. Create openclaw user
  2. Install nvm + Node 24
  3. npm install -g openclaw@latest
  4. scp ~/.openclaw/openclaw.json from Machine A
  5. openclaw daemon start (or onboard --non-interactive)
  6. openclaw doctor  ← validates everything

Key Points

Item	Detail
Agent capability	OpenClaw can run shell commands over SSH natively
Keys needed at install?	No — copy `openclaw.json` with existing config after bare install
Config location	`~/.openclaw/openclaw.json`
Full clone (including memory)	Copy entire `~/.openclaw/` + workspace, then run `openclaw doctor`
Security note	Never expose port 18789 to the public internet without authentication and TLS — use SSH tunnels for remote access

This is genuinely the most elegant way to do it — your agent handles the boring remote shell work and you end up with a clean clone of your setup.

References:

Back Donate