How to Deploy OpenClaw AI Assistant on a Home NAS (Step‑by‑Step)

Introduction

As a designer who enjoys tinkering with new tools, the author wanted a locally hosted AI assistant to avoid privacy and cost concerns of cloud services. OpenClaw, an open‑source AI assistant, was chosen for a fully controlled "digital butler" on a home server.

Preparation

Only a machine that can run 24 hours is required; a more powerful host is needed only for running large local models.

Hardware list

Small host : quiet and low‑power.

NAS (host) : Synology, FlyBox, or PVE – the author prefers PVE for flexibility.

Virtual system : Ubuntu 25 as the guest OS.

Good network : essential for a smooth installation.

Model selection

The author selected the MiniMAX model, the official recommendation for OpenClaw in China, and obtained a coding‑plan subscription for affordable usage.

Communication channel

Discord was chosen so family members can control the assistant via chat.

Environment Installation Notes

The installation steps are described sequentially.

Install Ubuntu in PVE

Any virtual machine works; the author assumes familiarity with PVE.

Set up password‑less sudo

OpenClaw requires elevated privileges, so password‑less sudo is configured. sudo visudo Add the following line at the end of the file (replace leolee with your username): leolee ALL=(ALL) NOPASSWD: ALL Save and exit:

With nano: Ctrl+O, Enter, then Ctrl+X.

With vim: type :wq and press Enter.

(Optional) Install SSH

Installing OpenSSH makes remote management smoother.

# Update package list
sudo apt update
# Install curl
sudo apt install -y curl
# Install and start OpenSSH server
sudo apt install -y openssh-server

Verify the service:

# Check status
sudo systemctl status ssh
# Start if not running
sudo systemctl start ssh

(Optional) Optimize network

A stable network, possibly with a global proxy on the router, is recommended.

Install OpenClaw Core

The author follows the official OpenClaw installation script:

curl -fsSL https://openclaw.ai/install.sh | bash

If the script fails, an alternative command can be tried: npm i -g openclaw && openclaw onboard After a successful install, run: openclaw onboard The configuration wizard appears. The author’s choices were:

Confirm the risk warning.

Select "QuickStart (Configure details later via openclaw configure.)".

Choose model provider "MiniMax".

Select "MiniMax M2.1" authentication method.

Enter the MiniMax API key.

Skip channel configuration for now.

Enable skill installation.

Enable all hooks.

Choose to open the Web UI.

Install shell completion script.

Apply the configuration:

source /home/leolee/.bashrc

(Optional) Start Web UI

To manage OpenClaw from other LAN computers, start the gateway: openclaw gateway --bind lan --port 18789 Set up SSH port forwarding (example IP 192.168.1.37):

ssh -N -L 18789:127.0.0.1:18789 [email protected]

Then launch the dashboard: openclaw dashboard Open the displayed address with token in a browser to access the Web UI.

Configure Communication Channel (Discord)

Run: openclaw config Select Discord; the terminal shows detailed instructions.

In the Discord developer portal, create an application, add a Bot, reset the token, and copy it back to OpenClaw.

Steps to generate a Bot invite link:

Open OAuth2 → URL Generator.

Check bot and applications.commands scopes.

In BOT PERMISSIONS, select message‑send, read‑history, or simply Administrator.

Copy the generated link, open it in a browser, and invite the Bot to your server.

Configure the Discord allowlist in OpenClaw by enabling Developer Mode, copying the server ID, and adding it to the Discord channels allowlist field.

Enable privileged intents:

PRESENCE INTENT

SERVER MEMBERS INTENT

MESSAGE CONTENT INTENT (most important)

In OpenClaw’s Web UI, turn on the Messages switch.

Fix Model Configuration Issues

When the Bot responded with errors, the author re‑ran: openclaw configure Selected Model → MiniMax → MiniMax M2.1 and re‑entered the API key. The correct model entry minimax-cn/MiniMax-M2.1 was chosen with the space key, then confirmed with continue.

If the Bot still failed, the baseurl in ~/.openclaw/openclaw.json was corrected from https://api.minimax.io/anthropic to https://api.minimaxi.com/, saved, and the gateway restarted:

openclaw gateway stop
openclaw gateway --bind lan --port 18789

After these fixes the assistant worked.

First Experiments

File control test:

在当前用户目录，创建一个md文档，写上hello world！

The assistant created the file smoothly.

Information retrieval test:

现在打开x首页，搜索ai资讯，找到前10条热门信息给我

The assistant returned search results, though more precise data would need a dedicated search API.

Integration with Moltbook community:

接入Moltbook，接入文档：curl -s https://moltbook.com/skill.md

The assistant guided the author through the integration, fetched documentation, and even posted a message on Moltbook.

Conclusion

The deployment journey—from hardware selection and virtual environment setup to step‑by‑step configuration—demonstrates the control and satisfaction of running a powerful AI assistant on a personal server, a experience that cloud‑only solutions cannot match. The author plans to add more skills such as automated design asset organization and intelligent layout suggestions.