marinabox 0.1.5

An open-source toolkit for deploying containerized desktops and browsers tailored for AI agents

MarinaBox

MarinaBox is a toolkit for creating and managing secure, isolated environments for AI agents. It provides:

Core Features

1. Secure Sandboxed Environments

   • Run isolated browser and desktop sessions locally or cloud
   • Perfect for AI agent tasks(Computer Use) and browser automation

2. Comprehensive SDK & CLI

   • Python SDK for programmatic control
   • Command-line interface for session management
   • Real-time monitoring and control capabilities
   • Integration with popular automation tools (Playwright, Selenium)

3. Interactive UI Dashboard

   • Live session viewing and control
   • Session recording and playback
   • Session management interface

Additional Features

• Cloud Integration: Deploy sandboxes to major cloud providers(coming soon)
• Multi-session Management: Run multiple isolated environments simultaneously
• Session Tagging: Organize and track sessions with custom tags

Documentation

Full documentation is available at https://marinabox.mintlify.app/get-started/introduction

Prerequisites

• Docker
• Python 3.12 or higher
• pip (Python package installer)

Installation

1. First, ensure you have Docker installed on your system. If not, install Docker for your operating system.

2. Pull the required Docker images:

docker pull marinabox/marinabox-browser:latest
docker pull marinabox/marinabox-desktop:latest

3. Install the Marinabox package:

pip install marinabox

Important Note

The provided Docker images are built for Mac ARM64 architecture (Apple Silicon). For other architectures:

1. Clone the sandbox repository:

git clone https://github.com/marinabox/marinabox-sandbox

2. Build the images with your target platform:

docker build --platform <your-platform> -f Dockerfile.chromium -t marinabox/marinabox-browser .
docker build --platform <your-platform> -f Dockerfile.desktop -t marinabox/marinabox-desktop .

Usage Example

Here's a basic example of how to use the Marinabox SDK:

from marinabox import MarinaboxSDK

# Initialize the SDK
mb = MarinaboxSDK()

# Set Anthropic API key
mb.set_anthropic_key(ANTHROPIC_API_KEY)

# Create a new session
session = mb.create_session(env_type="browser", tag="my-session")
print(f"Created session: {session.session_id}")

# List active sessions
sessions = mb.list_sessions()
for s in sessions:
    print(f"Active session: {s.session_id} (Tag: {s.tag})")

# Execute a computer use command
mb.computer_use_command("my-session", "Navigate to https://x.ai")

License

MarinaBox is released under the MIT License. See the LICENSE file for details.

Acknowledgments

This project makes significant use of:

• noVNC, an open source VNC client using HTML5 (WebSockets, Canvas). noVNC is licensed under the MPL-2.0 License.
• Anthropic Quickstarts, specifically the Computer Use Demo which provided inspiration for the sandbox implementation. Licensed under the MIT License.

Specify a custom video directory when initializing the SDK: ) Specify a custom video filename when stopping a session via SDK: ) 3. Use the CLI with custom video directory and filename: