> ## Documentation Index
> Fetch the complete documentation index at: https://noorle.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Sandbox

> Full compute environment with package installation. Run complex workloads with npm, pip, and system tools.

Sandbox provides a full Linux compute environment for running complex workloads. Install packages, use system tools, and execute long-running processes.

## Key Features

* **Full Linux Environment** - Ubuntu with bash shell
* **Package Management** - pip (Python), npm (Node.js)
* **System Tools** - curl, wget, git, ffmpeg, imagemagick, and more
* **Persistent Session** - Files persist within same session
* **Higher Limits** - 5-minute default timeout, up to 12GB memory

## How to Enable

### For Agents

1. **Agents** > Select Agent > **Settings** > **Capabilities**
2. Attach **Sandbox**

### For MCP Gateways

1. **Gateways** > Select Gateway > **Capabilities**
2. Attach **Sandbox**

## Usage Examples

### Install and Use Package

```bash theme={null}
#!/bin/bash
pip install requests beautifulsoup4
python3 << 'EOF'
import requests
from bs4 import BeautifulSoup

response = requests.get("https://example.com")
soup = BeautifulSoup(response.content, "html.parser")
print(len(soup.find_all("a")))
EOF
```

### Process Images

```bash theme={null}
#!/bin/bash
# Convert and resize image
convert input.jpg -resize 200x200 output.jpg
identify output.jpg
```

### Complex Data Pipeline

```bash theme={null}
#!/bin/bash
pip install pandas numpy scikit-learn

python3 << 'EOF'
import pandas as pd
import numpy as np

data = pd.read_csv("data.csv")
processed = data.dropna()
print(f"Rows: {len(processed)}")
print(processed.describe())
EOF
```

## Environment Details

### Available Languages

* **Python 3.11** - Full stdlib + pip
* **Node.js 20** - npm/yarn support
* **Bash/Shell** - Full command line access
* **Other tools** - git, curl, wget, ffmpeg, etc.

### Size Tiers

| Size         | vCPUs | Memory | Disk  |
| ------------ | ----- | ------ | ----- |
| XS           | 0.25  | 1 GB   | 4 GB  |
| SM (default) | 0.5   | 4 GB   | 8 GB  |
| MD           | 1.0   | 6 GB   | 12 GB |
| LG           | 2.0   | 8 GB   | 16 GB |
| XL           | 4.0   | 12 GB  | 20 GB |

### Storage

* **Session-scoped** - Data deleted after session ends
* **Files capability** - Persistent storage (separate)

## Resource Limits

| Limit                  | Value                     | Notes                         |
| ---------------------- | ------------------------- | ----------------------------- |
| Command Timeout        | 5 min default, 30 min max | Configurable per execution    |
| Max Lifetime           | 60 minutes                | Auto-terminates after 1 hour  |
| Idle Timeout           | 20 minutes                | Auto-terminates if idle       |
| Containers per Session | 1                         | One active at a time          |
| Default Size           | SM                        | 0.5 vCPU, 4 GB RAM, 8 GB disk |
| Network                | Outbound only             | No inbound connections        |
| Package Installation   | Unlimited                 | Any language packages         |

## Difference from Code Runner

| Feature      | Code Runner | Sandbox                   |
| ------------ | ----------- | ------------------------- |
| Timeout      | 30 seconds  | 5 min default, 30 min max |
| Memory       | 128MB       | 1 GB – 12 GB (by size)    |
| Storage      | None        | 4 GB – 20 GB (by size)    |
| Packages     | None        | pip, npm                  |
| System tools | None        | curl, ffmpeg, etc.        |
| Cost         | Low         | Higher                    |

**Use Code Runner for:** Quick calculations, simple tasks
**Use Sandbox for:** Package installation, heavy computation, external tools

## Common Use Cases

### Data Processing

```bash theme={null}
pip install pandas
python3 << 'EOF'
import pandas as pd
df = pd.read_csv("data.csv")
result = df.groupby("category").sum()
print(result)
EOF
```

### Web Scraping

```bash theme={null}
pip install beautifulsoup4 requests
python3 << 'EOF'
import requests
from bs4 import BeautifulSoup
# Scrape and parse HTML
EOF
```

### Image/Video Processing

```bash theme={null}
ffmpeg -i input.mp4 -vf scale=320:240 output.mp4
convert image.jpg -blur 0x8 blurred.jpg
```

### Database Operations

```bash theme={null}
pip install psycopg2 sqlalchemy
python3 << 'EOF'
import psycopg2
# Connect and query database
EOF
```

## Cost

For current pricing details, see [Pricing](https://noorle.com/pricing/).

Monitor in **Account** > **Usage** dashboard.

## Best Practices

### Install Strategically

Only install packages you need:

```bash theme={null}
pip install --no-deps package-name  # Skip dependencies if not needed
```

### Set Resource Limits

Prevent runaway processes:

```bash theme={null}
timeout 30 python3 script.py  # Auto-kill after 30 seconds
```

### Use Virtual Environments

Avoid conflicts with environment setup:

```bash theme={null}
python3 -m venv venv
source venv/bin/activate
pip install requirements
```

### Monitor Output

Keep track of what's happening:

```bash theme={null}
set -x  # Enable debug output
python3 script.py 2>&1 | head -100
```

## Troubleshooting

### Session Timeout

Command exceeded timeout (5 min default, 30 min max). Break into smaller tasks or optimize code.

### Out of Memory

Exceeded memory for your size tier. Scale up or process in batches.

### Package Installation Fails

* Check package name is correct
* Ensure compatible Python/Node version
* Try installing dependencies separately

### File Not Found

Session storage is temporary. Use Files capability for persistent storage.

## Next Steps

* [Code Runner - Quick Execution](/docs/run/capabilities/code-runner)
* [Files - Persistent Storage](/docs/run/capabilities/files)
* [Browser - Web Data Extraction](/docs/run/capabilities/browser)