# Overview

> Maindesk gives you and your AI agent the same computers in the cloud: Linux, Windows and a clean Mac. Your agent works there while you watch, answer and pick up the results from any browser.

Your agent gets a desk of its own: a Linux computer that keeps its files, with Windows and a Mac when the work needs them. You open the same desk on the web page, on a phone or a computer. Machines sleep when nobody uses them and count nothing asleep.

## How the pieces connect

```flow
You · Your agent → CLI · MCP · REST API → Your desk → Linux · Windows · Mac
You → The web page → Your desk
```

Your agent reaches the desk in one of three ways, all with the same key: the `maindesk` command line (CLI), MCP tools in its app, or the REST API. You see the same desk on the web page: a terminal for each machine, the files, the tasks and what needs you.

## What you can do

- **Hand long work to an agent.** A task runs on the desk, keeps going after you close your laptop, and sends a push when it starts, needs you or is done. See [Tasks](/docs/tasks).
- **Run commands on Linux, Windows or a Mac.** One call runs a command and returns its exit code and output. See [Machines](/docs/machines).
- **Build and test on a clean Mac.** Each Mac job starts on a fresh Mac and brings the files you name back to your desk.
- **Answer from your phone.** When an agent needs a login, a code or a choice, it asks you. You answer in the Inbox.
- **Share a result.** A finished task's page can be a read-only link that opens without sign-in.

## What I want to do

| What I want to do | Where |
| --- | --- |
| Get started from nothing | [Quickstart](/docs/getting-started) |
| Give an agent work that takes an hour | [Tasks](/docs/tasks) |
| Connect Claude Code, Codex or another agent on my computer | [Connect: CLI, MCP and API](/docs/connect) |
| Sign an agent in on my desk | [Agents](/docs/agents) |
| Know what a desk minute is and what a plan includes | [Plans and desk minutes](/docs/money) |
| Look up a command | [CLI commands](/docs/cli) |
| Call the API from a script | [API reference](/docs/api) |
| Understand an error | [Errors and fixes](/docs/errors) |
| Point my agent at the docs | [For AI agents](/docs/for-agents) |

## Next steps

- [Quickstart](/docs/getting-started): sign up, connect your agent and see your first task finish
- [Tasks](/docs/tasks): start, follow and answer tasks
- [Glossary](/docs/glossary): the words these docs use