MoltBot Security Guide: How to Sandbox & Prevent Data Loss

MoltBot is an incredibly powerful autonomous agent. To do its job, it needs access to your file system and terminal.

⚠️ WARNING: Never run MoltBot with root or Administrator privileges directly on your host machine. An AI agent can make mistakes. A simple hallucinated command like rm -rf / or overwriting your project's main.py could be catastrophic.

This guide will teach you how to create a Sandbox—a safe, isolated environment where MoltBot can work without risking your personal files.

🏆 Method 1: The Gold Standard (Docker)

The absolute safest way to run MoltBot is inside a Docker container. Even if the AI tries to delete everything, it can only delete files inside the container, leaving your actual computer untouched.

Step 1: Build the Image

First, navigate to the MoltBot folder and build the Docker image.

cd MoltBot
docker build -t moltbot .

Step 2: Run the Container

We will mount a specific folder (e.g., workspace) so the bot can save its work, but it cannot access anything outside this folder.

For Mac / Linux:

# Create a workspace folder first
mkdir -p $(pwd)/workspace

# Run container
docker run -it --rm \
  -v $(pwd)/workspace:/app/workspace \
  -p 3000:3000 \
  --name moltbot \
  moltbot

For Windows (PowerShell):

# Create a workspace folder
md workspace

# Run container
docker run -it --rm `
  -v ${PWD}/workspace:/app/workspace `
  -p 3000:3000 `
  --name moltbot `
  moltbot

Understanding the Flags:

| Flag | Purpose | |------|---------| | -v ...:/app/workspace | Maps your local folder to the container. The bot sees this as its whole world. | | --rm | Automatically deletes the container when you stop it (keeps your system clean). | | -p 3000:3000 | Allows you to access the dashboard in your browser. |

🥈 Method 2: Local Permission Boundaries

If you cannot use Docker, you must strictly limit where MoltBot can operate.

1. The "Playground" Rule

Never run MoltBot in your Home directory (~ or C:\Users\Name). Always create a dedicated "jail" folder.

mkdir ~/moltbot_playground
cd ~/moltbot_playground
# Run the bot ONLY from inside this folder
npm run start

2. Dedicated User (Linux/Mac Only)

Create a user with no sudo access and restricted file permissions.

# Create a user named 'moltbot-user'
sudo useradd -m moltbot-user

# Switch to this user before running the bot
sudo -u moltbot-user npm run start

If the bot tries to access system files, the OS will block it with "Permission Denied".

💰 Financial Security (API Costs)

Security isn't just about files; it's about your wallet. MoltBot runs loops and can consume API credits rapidly if it gets stuck.

1. Set Usage Limits

Do not rely on the bot to stop itself. Go to your API provider's dashboard and set a hard limit.

• OpenAI: Go to Billing limits. Set a "Hard Limit" (e.g., $10/month).
• Anthropic: Set a "Spend Limit" in your Usage Dashboard.

2. Protect Your .env File

Your API keys are as valuable as cash.

• Never commit .env to GitHub. Check your .gitignore file includes .env.
• Never show your .env file on a screen share.

# Check if .env is ignored
cat .gitignore | grep .env

Summary: Risk Levels

Which setup should you choose?

| Risk Level | Setup Method | Security Rating | |------------|--------------|-----------------| | 🟢 Low | Docker Container (Recommended) | ⭐⭐⭐⭐⭐ (Isolated filesystem) | | 🟡 Medium | Dedicated User (Linux/Mac) | ⭐⭐⭐ (OS-level permission blocks) | | 🟠 High | Local Folder | ⭐⭐ (Relying on AI to behave) | | 🔴 Critical | Running as Root/Admin | 💀 (Dangerous. Do not do this.) |

Our Advice

Spend the extra 5 minutes to install Docker. It is the only way to guarantee 100% safety.

Your documents and system files are worth more than a few minutes of setup time.