Skip to content
CLOUD COMPUTER

A real Linux desktop.
For your agent.

Start a computer for each of your users. Your agent sees the screen, clicks, types and runs commands. Your users watch it live in your app, or take over the mouse and keyboard.

Chrome, terminal and files Live view in your app Stops on its own when idle
CLOUD COMPUTER · LIVE
1280×800SEARCHTERMINAL$ ls ~Desktop  DownloadsexitCode 0$AGENTLOOK · ACTLIVE VIEWSCREENSHOTCLICKTYPEEXECPOST /computers/:id/actions
1280 × 800screen your agent sees and clicks
$0.10/hrfor Basic, counted by the second while running
$0while stopped, with every file kept
WHAT YOU GET

Everything included.
Nothing to wire up.

Linux desktops in the cloud for AI agents, one per user, live in your app.

A full desktop

Debian with Xfce, Google Chrome, Chromium, a terminal and a file system of its own, in an isolated sandbox.

One per user

Pass your id for a user as userId. Their computer keeps their files, browser sign-ins and screen, and asking again starts the same one.

Live in your app

Get a link to the live screen for an iframe. Let users watch the agent, or use the mouse and keyboard themselves.

Computer use built in

Screenshot, click, move, drag, scroll, type, keys and hotkeys, each one POST. Anthropic's and OpenAI's computer-use tools map straight onto them.

Run commands

The exec action runs a shell command and returns its exit code and output, so an agent can skip the mouse when a command is faster.

Stops when idle

With no mouse, keyboard or API activity, a computer stops by itself. Files stay, and nothing is charged while it is stopped.

HOW IT WORKS

Three steps
to production.

  1. 01
    Deploy

    Click Deploy on the Cloud Computer card. You get COMPUTERS_API_URL and COMPUTERS_API_KEY for your server.

  2. 02
    Start a computer

    POST /computers with your user's id, then show the preview link in an iframe.

  3. 03
    Run the agent

    Take a screenshot, send it to your model, and turn its answer into a click, keystroke or command. Repeat.

Read the Cloud Computer guide
agent.ts
// api() is a fetch to COMPUTERS_API_URL with COMPUTERS_API_KEY, as in the docs.

// Creates and starts this user's computer, or starts the one they have.
const computer = await api("/computers", { userId: user.id, size: "basic" });

// A link to the live desktop, for an <iframe> in your app.
const { url } = await api(`/computers/${computer.id}/preview`, { viewOnly: true });

const act = (action: object) => api(`/computers/${computer.id}/actions`, action);
const shot = await act({ type: "screenshot" }); // { image, mime, cursor }
await act({ type: "click", x: 640, y: 400 });
await act({ type: "type", text: "hello world" });
const out = await act({ type: "exec", command: "ls ~" }); // { exitCode, output }
BETTER TOGETHER

One project.
Every service connected.

Same project, same .env, same bill.

FAQ

Questions about
Cloud Computer.

What is on each computer?
A Linux desktop (Debian with Xfce) with Google Chrome as its default browser, Chromium, a terminal and its own file system, in an isolated sandbox.
Can my users take control?
Yes. A preview link with viewOnly: false lets them use the mouse and keyboard; viewOnly: true lets them only watch. Links work for an hour.
Does it work with Claude and OpenAI computer use?
Yes. Anthropic's and OpenAI's computer-use tools both work this way: turn each tool call into an action and send back the next screenshot.
How is it priced?
Only running time is charged, by the second, from your credits. Basic is $0.10 an hour. A stopped computer keeps its files at no charge, and idle computers stop by themselves.

Add Cloud Computer to your project.
Free to start.

Postgres, auth, storage, Redis, email and AI models. One API key.