How I Set Up a Secure, Sandboxed OpenClaw (Claudebot) for My Team |

Note: The tool mentioned in this article, formerly known as Claudebot, has been rebranded to OpenClaw. This article has been updated to reflect the new name and installation instructions. For the latest information, please visit the official website: openclaw.ai.

Many are rushing to buy Mac minis or spin up cloud servers for OpenClaw (formerly Claudebot), often leaving them exposed on the internet. I wanted to run OpenClaw on my computer, but not on my main system. I needed something sandboxed, something I could monitor closely. However, I didn’t want to buy a new machine, and I certainly didn’t want it on a server that could be compromised.

In this article, I’ll show you how I set up a sandboxed environment on my computer, installed OpenClaw inside it, and connected it through Slack for my entire team to use. The result? No exposed ports, no extra hardware, shared context, and a new AI employee for the whole team.

The Technology Stack

Here’s the setup we’ll be building:

Hyper-V: Windows’ built-in virtual machine software will serve as our sandbox. (Mac users have many similar options for running a VM sandbox).
Ubuntu: Inside the sandbox, we’ll run a lightweight and reliable Linux distribution.
OpenClaw: The AI agent will be installed within Ubuntu.
Slack: OpenClaw will connect out to Slack, allowing the entire team to interact with it securely without exposing any ports to the internet.

Step 1: Enable Hyper-V on Windows

First, we need to enable Hyper-V. It’s built into Windows as an optional feature, but this is true for Pro or Enterprise editions. I’m on Windows 11 Pro. If you’re on Windows Home, you’ll need to follow a different process.

The installation is quite simple. All you have to do is run a single command in PowerShell.

Open PowerShell as an administrator.
Copy and paste the following command:

Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Hyper-V -All

After the command runs, it will prompt you to restart your computer. Press Y and hit Enter. I have already installed it, so it’s not asking me, but this is a necessary step.

Once your computer restarts, you can search for “Hyper-V Manager” and run it.

Step 2: Create the Ubuntu Virtual Machine

In the Hyper-V Manager, you’ll see your computer’s name. You won’t see any existing virtual machines if this is your first time.

Click the Quick Create option. This opens a window with several operating system choices.
Initially, I tried the default Ubuntu option here but ran into roadblocks. A more reliable method is to download the OS image yourself.
Go to ubuntu.com and download the latest Ubuntu Desktop LTS version.
Back in Hyper-V’s Quick Create, choose Local installation source.
Crucially, untick the box that says “This virtual machine will run Windows.” We are installing Ubuntu.
Click “Change installation source,” navigate to your downloads, and select the Ubuntu ISO file you just downloaded.
Click Create Virtual Machine.

The basic setup is done in seconds. Before connecting, I wanted to fine-tune the machine’s resources.

Right-click the new VM and go to Edit Settings.
Under Memory, I increased the RAM from the default 4GB to 8GB for better performance.
I also disabled Dynamic Memory to prevent the VM from impacting my main system’s performance during day-to-day work.
Click OK.

Now, select the VM and click Connect.

Step 3: Install Ubuntu in the Sandbox

When you hit connect and start, your new sandboxed virtual computer will power up. We’ve only given it the installation disk (the ISO file), so now we need to go through the Ubuntu installation sequence.

The process is straightforward:

Just keep clicking “Next,” “Skip,” and “Install.”
I recommend checking two boxes when prompted:
- Install third-party software for graphics and Wi-Fi hardware
- Download and install additional media formats
From past experience, these options ensure a smoother experience.
You’ll see an option that looks scary: Erase disk and install Ubuntu. Don’t worry. This is erasing the sandbox’s virtual disk, not your main computer’s drive.
Proceed with the installation, set a name and password for your machine. Remember this password, as you’ll need it for admin access later.
The installer will detect your time zone. Click Install.

After a few minutes, your new virtual sandbox computer running Linux will be ready. It will ask to restart.

Note: My installation got stuck briefly on a message after the restart. I waited a couple of minutes, and when it didn’t proceed, I clicked the Reset icon in the Hyper-V toolbar. This resets the virtual machine, not your main computer. After that, it loaded perfectly.

Step 4: Install and Configure OpenClaw

Now for the main event. The rest of the setup requires copying and pasting several tokens. Moving them from my Windows host into the VM’s graphical interface can be annoying. A much easier way is to log into the virtual sandbox from my main computer’s terminal.

To do this, we’ll set up an SSH server in our sandbox.

In your Ubuntu VM, open a terminal.
First, update the package list:

sudo apt update

Enter your password when prompted.

Next, install the OpenSSH server:

sudo apt install openssh-server

Press Y to confirm.

Now, get the VM’s internal IP address:

ip addr

Look for an address under the eth0 interface, something like 172.30.99.39. This is the internal network address for communication between your Windows PC and the VM.

Open PowerShell on your Windows machine and connect via SSH. Replace youtube with your Ubuntu username and the IP address with the one you found:

ssh [email protected]

You’ll be asked to trust the connection; type yes. Enter your password, and you are now logged into your virtual sandbox from PowerShell. This makes copy-pasting much easier.

Now, let’s install OpenClaw.

Go to openclaw.ai and find the installation command for macOS/Linux. It should be a curl command.
Copy the command, paste it into your PowerShell window (which is connected to the VM), and hit Enter.

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

This script will install OpenClaw and any dependencies it needs, then start the onboarding experience.

Step 5: Onboarding and API Configuration

The setup will guide you through a few prompts.

Agree to the terms.
Choose Quick start for the onboarding mode.
Select your AI provider. I’m using Anthropic (Claude), but you can choose from others.
The setup will now ask for an API token.

To get the token, you need to run a command on your host machine where you have your primary OpenClaw setup.

Open a new, separate PowerShell window on your Windows PC.
Run the command: openclaw setup token.
This opens a browser window to authorize the connection. Click Authorize.
Back in PowerShell, a token will be generated. Copy this token.
Go back to the PowerShell window connected to your VM.
Confirm you are using an Anthropic token, paste the token you just copied, and press Enter.
You can accept the defaults for the token name and model.

Step 6: Connecting OpenClaw to Slack

Next, select the channel. I’m installing this for Slack so the whole team can access it securely.

Choose Slack (socket mode). This is a secure method that doesn’t require opening ports.
The setup provides a link to the best guide in the OpenClaw documentation for this process.

Here’s a summary of the steps:

Create a Slack App:
- Go to api.slack.com/apps and click Create New App.
- Choose From scratch.
- Give it a name (e.g., “OpenClaw”) and select your workspace.
- Click Create App.
Use the App Manifest:
- In the OpenClaw documentation, find the optional manifest file. It’s a block of YAML configuration.
- Copy the entire manifest.
- In your Slack app’s settings, go to App Manifest, delete the existing content, and paste the manifest you copied.
- Save Changes. This automatically configures all the necessary settings, including socket mode.
Install the App and Get Tokens:
- Go to Install App and install the app to your workspace.
- You will be presented with a Bot User OAuth Token (starts with xoxb). Copy this.
- Paste the xoxb token into the OpenClaw setup prompt in your terminal.
Generate an App-Level Token:
- Next, the setup asks for a Slack App Token (starts with xapp).
- In your Slack app settings, go to Basic Information, scroll down to App-Level Tokens, and click Generate Token and Scopes.
- Give the token a name (e.g., “openclaw-app-token”).
- Add the scope connections:write.
- Click Generate.
- Copy the generated xapp token.
- Paste this token into the OpenClaw setup prompt.

Step 7: Finalizing the Setup

You’re almost there!

Configure Slack channel access and skills as prompted. For now, you can accept the defaults and skip adding extra skills.
When asked about Hooks, I recommend enabling all three to log commands and save session memory. This is part of OpenClaw’s power.
The installer will set up the gateway service.
When asked to launch the TUI or web, choose TUI (Terminal User Interface).

OpenClaw is now running! You can give it a name (like Jarvis) and set its personality.

Step 8: Pairing with Slack

The final step is to pair your Slack user with the running OpenClaw instance.

Open the Slack workspace where you installed the app. You should see the OpenClaw app.
Send it a message, like “Are you here?”.
It will respond with a message saying access is not configured and provide a pairing code.
Go back to your terminal running the OpenClaw TUI.
Type /exit to leave the chat interface and return to the OpenClaw command line.
Type the following command, replacing the code with the one from Slack:

openclaw pairing approve slack GKB78XY

OpenClaw is now approved and connected. Go back to Slack and send another message. Jarvis should now respond, continuing the same conversation you started in the terminal.

You now have OpenClaw running securely on your own machine, with your whole team able to access it through Slack. No security worries, shared context across the team, and a powerful new AI assistant at your service.