Quickstart

This quickstart will walk you through how to build a simple agent with an Archil disk connected to the Vercel AI SDK. Archil supports multiple ways to use disks. If you’re looking to learn how to mount a disk to a server or sandbox that you’re already running, check out our documentation for Linux or Mac. If you want to see how to use Archil with our serverless sandboxes, then checkout Serverless sandboxes.

This quickstart relies on features which are only available in Archil’s AWS regions.

Get an API key

Sign in to console.archil.com, and create an API key from the API keys page. Create an API key in the region where you want your Archil disk to be stored. This should ideally be the same region as your application.Export these variables:

export ARCHIL_API_KEY=key-...
export ARCHIL_REGION=aws-us-east-1

Create a disk

Create a disk. With no data source attached, Archil tiers its data to storage it manages for you — nothing to provision.

npx disk create quickstart-disk

If you want to give your agent access to data that already lives in an S3 or GCS bucket, you can instead create a disk with a data source — see how in S3 Object Storage.

Populate your disk with some data

Run a command like an agent

disk exec runs a bash command in a container with the file system already mounted, and returns stdout, stderr, and the exit code:

npx disk my-disk exec "ls -la"

Try something more interesting — summarize the bucket, or pipe a result into wc -l:

npx disk my-disk exec "du -sh ."
npx disk my-disk exec "find . -name '*.json' | wc -l"

No data is copied to your laptop. Only the result comes back. Billing is the wall-clock time the command runs (1ms increments, 100ms minimum). See Serverless sandboxes for the full model and the disk CLI reference for every command and flag.To search every file in the bucket for a pattern, use the purpose-built Search Files primitive (disk.grep) from the SDK instead of shelling out to grep — it fans the search across many containers in parallel.

Or do it from code

The same flow from TypeScript or Python:

import * as archil from "disk";

const { disk } = await archil.createDisk({
  name: "my-disk",
});

const { stdout } = await disk.exec("ls -la");
console.log(stdout);

import archil

result = archil.create_disk(name="my-disk")

res = result.disk.exec("ls -la")
print(res.stdout)

See the TypeScript SDK reference or the Python SDK reference for the full API.

Next steps

Mount to a Linux server

Run on Lambda or a worker

Use disk.exec from AWS Lambda or Cloudflare Workers without provisioning compute.

Build a bash tool for an agent

Wire disk.exec into an AI agent loop — about 30 lines.

Pricing

View Archil pricing and plan details.

Getting started

Compute

Concepts

Data Sources

Details

Administration

Protocols

Reference

Legal

Next steps

Mount to a Linux server

Run on Lambda or a worker

Build a bash tool for an agent

Pricing

​Next steps

Mount to a Linux server

Run on Lambda or a worker

Build a bash tool for an agent

Pricing

Next steps