Introduction

• Run untrusted AI-generated code in sandboxes. Each persistent sandbox is a Firecracker microVM with its own Linux kernel, a persistent root disk, and controlled egress. For one-off commands against a disk, use serverless execution.
• Build and give agents access to petabyte-scale context, using our infinite, shared disks and optionally synchronizing that data from sources like Amazon S3 object storage.
• Share artifacts between agents and humans with a strongly consistent REST API that allows you to easily embed file watching and uploading into any website.

import { generateText } from "ai";
import * as archil from "disk";
import { createDiskTools } from "@archildata/ai-sdk";

const disk = await archil.getDisk('disk-1234');
const result = await generateText({
    model: "anthropic/claude-sonnet-4.6",
    prompt: "What files do you have access to?",
    tools: createDiskTools(disk),
});

import * as archil from "disk";

const myDisk = await archil.getDisk('disk-1234');

// Easily hand off between bash and REST
//
await myDisk.exec('curl google.com > google.txt');
console.log(await myDisk.getObject('google.txt'))

​

Features

• Branches and checkpointing make it possible to have simultaneous agents run on the same data set, and pick the best output.
• Native SQLite support allows you to easily create databases on your file system, and serverlessly read and write to them from anywhere.
• Immutable log support simplifies how you build log-like abstractions on your disk.

​

Pricing