Overview

​

Core Capabilities

• Sandbox Management: Create, configure, and manage isolated computing environments with resource limits and persistent state.
• Code Execution: Execute Python, JavaScript, and shell commands safely with full library support and context management.
• File Operations: Upload, download, read, write, and organize files within sandboxes with complete file system access.
• Templates: Pre-configured and custom environments with popular libraries and tools for different programming languages and use cases.
• LLM Integration: Connect sandboxes with Large Language Models to build powerful code-interpreting agents and data analysis tools.

​

Multi-Cloud Support

​

Available Regions

​

Common Use Cases

• Run AI-Generated Code: Safely execute code produced by LLMs and AI agents in isolated sandboxes with no risk to your infrastructure.
• Code Interpreter: Build ChatGPT-style code interpreter experiences — let users write, run, and iterate on code in real time.
• Deep Research: Power autonomous research agents that browse, scrape, analyze data, and synthesize findings across multiple steps.
• AI Evaluations: Run evaluation harnesses and benchmarks in isolated environments with reproducible results.
• Data Analysis & Visualization: Process datasets, run statistical analysis, and generate charts and reports programmatically.
• Automated Workflows: Execute scripts, data processing pipelines, and automation tasks in controlled, reproducible environments.

​

Getting Started in 3 Steps

1. Install the SDK: Install the GravixLayer Python SDK.
2. Create a Sandbox: Choose a template to create your isolated execution environment.
3. Execute Code: Run Python, JavaScript, or shell commands with persistent state and context management.

from gravixlayer import GravixLayer

# Initialize the client (defaults to azure / eastus2)
client = GravixLayer(api_key="YOUR_API_KEY")

# Create a sandbox from a template
sandbox = client.sandbox.sandboxes.create(
    template="python-base-v1",
    timeout=300,  # 5 minutes
)

print(f"Sandbox ID: {sandbox.sandbox_id}")
print(f"Status: {sandbox.status}")

# Run code in the sandbox
result = client.sandbox.sandboxes.run_code(
    sandbox.sandbox_id,
    code="print('Hello from AgentBox!')",
    language="python",
)
print(result.logs.stdout)

# Clean up
client.sandbox.sandboxes.kill(sandbox.sandbox_id)