Copy page

Get started

Quickstart

From sign-up to your agent's first click in about five minutes.

1. Get a screen

Sign in at https://console.trykubi.com and buy your first screen, or top up credit and make a disposable one for a short run (two kinds). It is ready in seconds.

2. Create an API key

In the dashboard, open API keys, create a key and copy it. It is shown once. Give it the read and control permissions for this quickstart.

Code
export KUBI_API_KEY=kubi_...

3. Install the SDK

Code
npm install @trykubi/sdk

4. Look, then act

Code
import { Kubi } from "@trykubi/sdk";

const client = new Kubi(); // reads KUBI_API_KEY
const screen = client.screen();   // your only screen; pass an id when you have several

const shot = await screen.screenshot();
console.log(shot.width, shot.height, shot.screenshot_id);

const result = await screen.click({
  x: 640,
  y: 360,
  screenshot_id: shot.screenshot_id, // refuse to click if the screen changed
  screenshot: "after",              // get the screen back once it settles
});
console.log(result.status, result.screenshot?.screenshot_id);

The same two calls with curl:

Code
curl https://api.trykubi.com/v1/screens/scr_7f2kq9w3x8m4b6tzp1ac/screenshot \
  -H "Authorization: Bearer $KUBI_API_KEY"

curl -X POST https://api.trykubi.com/v1/screens/scr_7f2kq9w3x8m4b6tzp1ac/click \
  -H "Authorization: Bearer $KUBI_API_KEY" \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{"x": 640, "y": 360, "screenshot": "after"}'

5. Watch it live

Open the screen in the dashboard to see the desktop as your agent works. Press Take control and confirm to use it yourself. Your agent's calls are refused until you press Release or 5 minutes pass with no input from you.

Next

• Give an agent app the screen instead of writing code: Connect your agent
• How looking and acting fit together: The look and act loop
• Every endpoint: API reference