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

# bead machine concepts

> Learn bead machine concepts to choose machine kinds, templates, snapshots, forks, fleets, swarms, and worktrees.

Learn bead machine concepts for Linux commands, files, and optional desktop access. The CLI, SDKs, MCP, and dashboard use the same account resources.

## Lifecycle

Create a machine from a template, wait for readiness, and run work. The machine's kind determines what happens when it is idle or its runtime ends.

| Kind | Behavior |
| - | - |
| Temporary | Runs until its timeout, then deletes. Maximum timeout is 24 hours. |
| Persistent | Keeps memory and disk across pause. Pauses when idle and resumes with funding. |

Persistent machines default to a 15-minute idle timeout. Paused machines incur retained storage charges. Read [Machine lifetime](https://docs.bead.run/run-work/lifetime.md) for funding and expiry details.

## Words

| Term | Meaning |
| - | - |
| Machine or sandbox | Your Linux execution environment. API paths often call it a sandbox. |
| Size | CPU, memory, and disk capacity selected when creating a machine. |
| Template | A published starting image or prepared snapshot. |
| Snapshot | A saved capture of machine memory, disk, and device state. |
| Fork | An independent copy of captured machine state with the original size. |
| Fleet | A batch of machines created together, with a result for each member. |
| Swarm | A command distributed over temporary machines, with collected results. |
| Worktree | A local Git worktree bound to its own persistent remote machine. |
| Organization | The account boundary for machines, credentials, membership, and credit. |
| Prepaid credit | Funds used to reserve and settle work and retained storage. |

## Choose a starting point

Use [a temporary job](https://docs.bead.run/run-work/run.md) for a bounded command, [worktrees](https://docs.bead.run/worktrees/overview.md) for local development, and [fleets or swarms](https://docs.bead.run/use-cases/parallel-work.md) for parallel work.

A memory snapshot or fork keeps the source CPU and memory bounds. To choose a different size, create a new machine from a disk template. See [Machine sizes](https://docs.bead.run/machines/sizes.md).

## Pay for what you use

CPU uses measured busy time. Memory uses configured running memory. Storage uses unique retained bytes. Check [workload estimates](https://docs.bead.run/platform/estimates.md), [balance](https://docs.bead.run/platform/prepaid-credit.md), and [effective limits](https://docs.bead.run/platform/limits.md) before provisioning.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.