> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sailresearch.com/llms.txt
> Use this file to discover all available pages before exploring further.

# What are Sailboxes?

> Efficient cloud environments for long-horizon agents

This page is a high-level guide to Sailboxes, efficient cloud environments
designed for rollouts and long-horizon agents.

See the [Sailbox reference](/sailbox-sdk) for reference documentation on
Sailboxes.

## What are Sailboxes and why should I use them?

Sailboxes are persistent Linux VMs designed for long-horizon agents. They can be
dynamically provisioned in seconds and are the perfect cloud environment for any
agent.
They provide a number of advantages over other sandboxing providers:

* Cost efficiency: we are both >70% cheaper than other providers and
  only charge you for the exact portion of CPU, memory, and disk you
  use. Since agents spend most of their time blocked on I/O, we are
  significantly more cost-efficient than anyone else.
* Elastic scaling: use as much (or as little) compute as you need, up to your
  Sailbox size's CPU, memory, and disk ceilings. Your agent will never OOM, and
  because billing follows actual usage, you will never be charged for unused capacity.
* Pause and resume all sandbox state. In-flight work survives the pause
  with no save/restore code on your side, and open network connections survive
  for up to 10 minutes.

## Provider comparison

<div className="sailboxes-comparison-table">
  | Provider      | vCPU                          | Memory                       | Sleep during inference | Max runtime    | Memory snapshots | Local NVMe disks | Docker-in-Docker | Start/resume time |
  | ------------- | ----------------------------- | ---------------------------- | ---------------------- | -------------- | ---------------- | ---------------- | ---------------- | ----------------- |
  | Sailboxes     | \$0.015<br />active vCPU·h    | \$0.008<br />active GB·h     | Yes                    | No fixed limit | Yes              | Yes              | Yes              | \<2s              |
  | Modal         | \$0.071<br />active vCPU·h    | \$0.008<br />reserved\* GB·h | No                     | 24 hours       | Alpha            | No               | Beta             | \<500ms           |
  | E2B           | \$0.0504<br />reserved vCPU·h | \$0.0162<br />reserved GB·h  | No                     | 24 hours       | Yes              | No               | Yes              | \<1s              |
  | Vercel        | \$0.128<br />active vCPU·h    | \$0.0212<br />reserved GB·h  | No                     | 5 hours        | No               | Yes              | Yes              | \<1s              |
  | Daytona       | \$0.0504<br />reserved vCPU·h | \$0.0162<br />reserved GB·h  | No                     | No fixed limit | Experimental     | Yes              | Yes              | \<500ms           |
  | Cloudflare    | \$0.072<br />active vCPU·h    | \$0.009<br />reserved GB·h   | No                     | No fixed limit | No               | Yes              | Yes              | \<3s              |
  | AWS AgentCore | \$0.0895<br />active vCPU·h   | \$0.00945<br />active GB·h   | No                     | 8 hours        | No               | No               | No               | 1s                |
</div>

<p className="sailboxes-comparison-footnote">
  \*Modal allows bursting beyond reserved capacity, but this is best-effort
  (e.g. sandboxes OOM if no memory is available).
</p>

## What is the tradeoff?

Sailboxes achieve their efficiency through live migrations: we are able to
achieve much higher utilization on our underlying hardware than other providers
by migrating VMs based on their live resource usage. These migrations only occur
a few times a day, take several seconds, and happen completely without the
knowledge of the agent.

While these migrations mean that the occasional command incurs a few seconds of
additional latency, the tradeoff is massive gains in cost-efficiency. Unlike
human-in-the-loop workflows like chatbots where patience is limited,
long-horizon agents can tolerate the occasional latency blip without issue.
Sailboxes directly trade off this p99 latency for better economics and we
believe this makes them the ideal platform for any long-horizon agent.

## Start here

* [Quickstart](/sailboxes-quickstart): create a Sailbox, run commands, expose a
  service, and clean up.
* [Migrate to Sailboxes](/sailboxes-migrating): move an app, an agent, or a
  dev environment onto Sailboxes, with your coding agent doing the work.
* [Pricing](/sailboxes-pricing): understand observed-usage billing dimensions and
  per-hour rates.
* [Network policy](/sailboxes-network-policy): limit what a Sailbox can reach,
  or cut it off from the network entirely.
* [Credential injection](/sailboxes-credentials): let a Sailbox call
  authenticated HTTPS APIs without storing credentials inside it.
* [Harbor](/harbor): run Harbor evaluation tasks on Sailboxes with one flag.