Local Sandbox
What is Local Sandbox?
Local Sandbox is a localized runtime environment provided by 302 AI Studio that deploys Agents on your local computer through Docker container technology. Compared to cloud sandbox, local sandbox gives you faster response times and stronger data control.
Remote Sandbox vs Local Sandbox
☁️ Remote Sandbox
No Configuration Required, Ready to Use
- Zero configuration, instant start
- Cloud-based, no local resources needed
- Automatic maintenance and updates
- Suitable for quick experience and lightweight usage
💻 Local Sandbox
Local Execution, Ultra-Fast Response
- Local deployment, data stays local
- Faster response, no network latency
- Full control over runtime environment
- Ideal for scenarios requiring higher performance and privacy
Why Use Containers Instead of Directly Using Claude Code?
- Sandbox Isolation - Containers act as an independent sandbox environment where all AI operations are restricted within the container. Even if AI code has issues or performs dangerous operations, it won't affect your real system
- Directory Mounting - Our working directories are mapped into the container via mounting. You can freely choose which folders to share with AI, while other folders remain completely invisible to AI, providing better privacy protection
- Permission Control - The AI execution environment within the container has restricted permissions and cannot access your sensitive files, system configurations, or network credentials
- Environment Consistency - Containers provide a uniform runtime environment, ensuring AI code runs the same way on any machine, avoiding "works on my machine" problems
- Resource Limits - Precisely control the CPU, memory, and disk resources that AI tasks can use, preventing runaway code from consuming all your system resources
- Easy Cleanup - All AI execution traces, temporary files, and installed dependencies are contained within the container, which can be quickly and completely cleaned up when you close the client, leaving no residue
How It Works
Local Sandbox is implemented through the following technologies:
- Containerized Deployment: Uses Podman as the container engine to run pre-packaged Agent service images
- Local Service: Agent service runs locally and is accessed via localhost
- Automatic Management: Client automatically manages container startup, shutdown, and resource allocation
Use Cases
Local Sandbox is particularly suitable for the following scenarios:
| Scenario | Description |
|---|---|
| High-Frequency Usage | Requires frequent use of AI features with pursuit of ultimate response speed |
| Data Privacy | Strict requirements for data privacy, wanting data to stay local |
| Offline Usage | Use in environments with unstable networks or requiring offline work |
| Development & Debugging | Developers need to locally debug and test AI features |
Important Notes
Local Sandbox requires installing Podman and Docker Compose on your computer. Container images will be automatically downloaded on first startup. Please ensure you have sufficient disk space (at least 10GB recommended) and a stable network connection.
Local Sandbox will occupy certain system resources (CPU, memory, disk). It is recommended to use on computers with higher specifications. Recommended configuration: 8GB+ RAM, 4+ CPU cores.