Getting Started with AgentOS: A Hands-On Tutorial

Getting Started with AgentOS: A Hands-On Tutorial

In this tutorial, we'll walk through setting up AgentOS and creating your first AI agent with file system access. AgentOS changes the game by giving your AI agents a lightweight, secure operating system that runs as a library — no Docker, no VMs, no SaaS required.

Prerequisites

• Node.js 18+ or Bun
• Basic familiarity with JavaScript and AI agents
• 2GB RAM minimum (4GB recommended for production)

Step 1: Installation

Install AgentOS via npm:

npm install @rivet-ai/agentos

Step 2: Create Your First Agent

Here's a simple example that creates an agent with file system access:

import { AgentOS } from '@rivet-ai/agentos';

const os = new AgentOS({
  storage: './agent-data',
  memory: 256, // MB
  network: true
});

const agent = await os.spawn({
  name: 'my-first-agent',
  runtime: 'wasm',
  entrypoint: './agent.wasm'
});

await agent.fs.writeFile('/data/hello.txt', 'Hello from AgentOS!');
const content = await agent.fs.readFile('/data/hello.txt');
console.log(content); // Hello from AgentOS!

await os.shutdown();

Step 3: Multi-Agent Setup

One of AgentOS's superpowers is running multiple agents concurrently. Each agent gets its own isolated OS instance:

const agents = await Promise.all(
  tasks.map(task => os.spawn({
    name: `agent-${task.id}`,
    runtime: 'wasm',
    context: { task }
  }))
);
const results = await Promise.all(
  agents.map(agent => agent.waitForCompletion())
);

Performance Benchmarks

Conclusion

AgentOS is a paradigm shift for AI agent development. By replacing heavyweight containers with WASM-based isolates, it enables a new class of multi-agent applications that were previously impractical. Whether you're building code generation tools, autonomous research assistants, or data processing pipelines, AgentOS gives you the operating system your agents deserve.