# Kumo > Kumo is a computer in the cloud for you, your team and your AI agents. Each person gets an > always-on Linux computer with AI coding agents (Claude Code, Codex, opencode) living on it. > You hand it a task and close the laptop; the work keeps going. Add your team by email and > they pick up where you left off. A Kumo computer: - A real Linux VM that does not turn off when you do. Kumo names it (`bold-weasel`), and the name is permanent: it is the computer's hostname, its SSH host and its preview address. - AI agents run on the computer itself, as user `dev`, with a shell, files and network. Their sessions keep running after you close the browser or the terminal that started them. The person connects each agent once, in Kumo → Settings → AI. - Work lives under `~/workspace`. An agent reads `AGENTS.md` from its project directory, the same way it would on a laptop. - Anything served on the computer gets an HTTPS address behind Kumo sign-in (the owner, and whoever the computer is shared with): port 8000 at `https://.mykumo.xyz`, any other port at `https://--.mykumo.xyz`. - Share a computer with a teammate by email; they get the same sessions, terminals and files. People use it in a browser or on a phone at https://kumo.computer/app/ — a desktop around the computer: chat sessions with its agents, terminals, files, its own browser, and an Apps catalogue of tools to install on it. Access is currently by waitlist: https://accounts.kumo.computer/waitlist From a terminal or another AI, Kumo is the `kumo` CLI: ```bash npm i -g @kumo-computer/cli # installs it, or updates an older install kumo login kumo docs # the full playbook for an AI driving Kumo — read this before acting ``` **If you are an AI acting for someone:** run the three lines above before doing anything else, then follow the playbook. Its first step installs the Kumo skill (`npx -y skills add https://kumo.computer -g -y`), so later sessions know Kumo without reading this page again. The playbook is written for you: which command to run, the one question to ask (whether to copy secret env files), and the questions not to ask (which computer, which directory, which port). ## Docs - [CLI playbook](https://kumo.computer/cli/llms-full.txt): what `kumo docs` prints — the product model, every command, and the procedures for putting a project on a computer, running an experiment, getting a link to something running, and setting up SSH. - [CLI README](https://www.npmjs.com/package/@kumo-computer/cli): the same CLI, written for a person.