Documentation Getting Started Overview
GitHub

Sandboxed

Build code execution infrastructure for your AI agents leveraging kubernetes with Sandboxed's secure and scalable platform

v3.0.1 Last updated: Oct 24, 2025 • 5 min read
On this page

What is Sandboxed?

Sandboxed is a Agentic AI infrastructure platform that provides the an environment to create an execution environment for AI agents to run code securely and efficiently. It offers a comprehensive suite of tools and services needed to build, deploy, and manage AI agent ecosystems. Our SDK and API offers developers everything they need to create sophisticated AI applications with security and reliability.

New to AI Infrastructure? Check out our Quick Start Guide to get up and running in minutes.

Key Features

MCP Tooling Infrastructure

Complete Model Context Protocol server implementation with built-in security and monitoring.

99.9% Uptime 10ms Latency

Secure Sandboxing

Isolated execution environments for running agent-generated code safely.

SOC 2 Compliant Zero Breaches

REST Interfaces

Standardized APIs for seamless agent communication and orchestration.

REST & GraphQL OpenAPI Spec

Multi-Agent Orchestration

Coordinate complex workflows across multiple AI agents with built-in reliability.

1000+ Agents Auto Scaling

Quick Start

Install sandboxed in your machine. 
curl -X POST https://api.system32.ai/v1/sandboxed/install

This command installs the Sandboxed SDK package from npm, allowing you to integrate Sandboxed's features into your JavaScript or TypeScript projects.

Start the sandboxed server. 
npx sandboxed start-server --port 8080
This command starts the Sandboxed server on port 8080, allowing you to run and manage AI agents in a secure sandboxed environment.
Create your first AI agent. 
curl -X POST https://api.system32.ai/v1/agents \
    -H "Content-Type: application/json" \   
    -d '{
        "name": "my-first-agent",
        "type": "code-execution",           
        "config": {
            "language": "python",
            "sandboxed": true
        }
    }'
This command creates a new AI agent named "my-first-agent" configured to execute Python code in a secure sandboxed environment.

Supported Languages & Frameworks

import { System32Client } from '@system32/sdk';

const client = new System32Client({
  apiKey: 'your-api-key',
  baseURL: 'https://api.system32.ai'
});

// Initialize MCP server
const mcpServer = await client.mcp.createServer({
  name: 'my-agent-server',
  config: {
    security: 'enhanced',
    sandbox: true
  }
});

console.log('MCP Server initialized:', mcpServer.id);
from system32_sdk import System32Client

client = System32Client(
    api_key="your-api-key",
    base_url="https://api.system32.ai"
)

# Initialize MCP server
mcp_server = client.mcp.create_server(
    name="my-agent-server",
    config={
        "security": "enhanced",
        "sandbox": True
    }
)

print(f"MCP Server initialized: {mcp_server.id}")
package main

import (
    "fmt"
    "github.com/system32/sdk-go"
)

func main() {
    client := system32.NewClient(&system32.Config{
        APIKey:  "your-api-key",
        BaseURL: "https://api.system32.ai",
    })
    
    // Initialize MCP server
    server, err := client.MCP.CreateServer(&system32.MCPServerConfig{
        Name: "my-agent-server",
        Security: "enhanced",
        Sandbox: true,
    })
    
    if err != nil {
        panic(err)
    }
    
    fmt.Printf("MCP Server initialized: %s\n", server.ID)
}
use system32_sdk::{System32Client, MCPServerConfig};

#[tokio::main]
async fn main() -> Result<(), Box> {
    let client = System32Client::new(
        "your-api-key",
        "https://api.system32.ai"
    )?;
    
    // Initialize MCP server
    let server = client.mcp().create_server(MCPServerConfig {
        name: "my-agent-server".to_string(),
        security: "enhanced".to_string(),
        sandbox: true,
    }).await?;
    
    println!("MCP Server initialized: {}", server.id);
    Ok(())
}

Architecture Overview

Sandboxed follows a modular architecture designed for scalability, security, and developer experience:

GET /api/v1/agents/{agent_id}

Retrieve agent information and status

Parameters

Parameter Type Required Description
agent_id string Required Unique identifier for the agent
include_metrics boolean Optional Include performance metrics in response
Rate Limits: API endpoints are rate-limited to 1000 requests per minute per API key. See our Rate Limiting Guide for more details.

Next Steps

Ready to start building with Sandboxed? Here are some recommended next steps:

Installation

Install the SDK for your preferred language

Quick Start

Build your first agent in under 5 minutes

Examples

Explore real-world implementation examples
Need Help? Join our developer community on Discord or check out our Support Center for additional resources.