# Set up screenRIG in Gemini Spark.

> Canonical HTML: https://screenrig.ai/docs/start/gemini-spark/

Install the official screenRIG plugin inside a Gemini Spark task, connect your screenRIG
account, pair a Player, and let Spark publish your first screen. Spark is Google’s
personal AI agent inside Gemini Apps: you work in a task thread, not a terminal.

Spark is **not** Gemini CLI, and this guide does not claim a tested Spark integration.
Spark’s task execution is Google’s infrastructure; the steps below verify that your
Spark environment can actually run the official plugin’s bundled skill and CLI before
you rely on it.

## Get Gemini Spark

1. **Open Gemini Spark.** Start at Google’s
   [Spark overview](https://gemini.google/overview/agent/spark/), then open the
   Gemini app and choose **Switch to Spark**. In the
   [Gemini app on Mac](https://gemini.google/mac/), choose **Spark → Tasks**.
2. **Check availability.** Follow
   [Google’s Spark help article](https://support.google.com/gemini/answer/17094507?hl=en)
   for current account requirements, subscriptions, platforms, and country availability.

Your Google account is separate from your screenRIG account. Keep them apart: Spark
never needs your screenRIG password or API tokens.

## Ask Spark to install screenRIG

Start a Spark task with an explicit authorization:

> I authorize you to install the official screenRIG plugin from https://github.com/screenrig/plugin.

The official plugin bundles the screenRIG skill and the CLI your agent uses to operate
screens; its CLI needs Node.js 20.11 or newer. Use the task prompt above rather than
the Claude Code, Codex, or Grok marketplace commands: those commands are not for Spark.
Review the installation and tool-access approvals Spark requests in the task thread.

## Check the execution environment

A Spark reply describing success is not proof. Ask it to show the command output:

```console
screenrig version
screenrig doctor
```

For an installed plugin, follow Start’s
[session PATH setup](https://screenrig.ai/docs/start.md#run-the-bundled-cli).
Ask Spark to run `node --version` if `version` fails on a Node error.

Spark’s remote code execution and saved remote files live on Google’s remote computer,
not on your machine. If that environment cannot run the bundled CLI — no shell, no
Node.js 20.11 or newer, or blocked commands — stop and ask which capability is missing.
Continuing from a chat reply alone leaves the plugin uninstalled. In that case use
[Claude Code](https://screenrig.ai/docs/start/claude-code/) or
[ChatGPT Codex](https://screenrig.ai/docs/start/chatgpt/), which run the same plugin on
your own machine.

<a id="account"></a>

## Connect your screenRIG account

For a new account, give Spark your actual contact email so it can run
`screenrig agent enroll --email ADDRESS`. If you already have a screenRIG account, say
so and complete the dashboard approval Spark requests for `screenrig agent connect`.
Never paste your password, account credentials, or API tokens into the task thread —
Google’s own Spark guidance says not to enter sign-in or payment details directly in a
task. Let Spark open the page and sign in yourself if it asks.

Both paths in detail: [Account access](https://screenrig.ai/docs/start.md#account-access).

<a id="player"></a>

## Pair your Player and publish

1. **Open a Player.** On the device driving your screen, open
   [play.screenrig.ai](https://play.screenrig.ai) or follow
   [Player downloads & setup](https://screenrig.ai/docs/players.md). Use the Player web app
   (PWA), the available customer Player today. Send its pairing code to Spark so it can run
   `screenrig screen pair CODE` and confirm the intended screen with `screenrig screen list`.
2. **Describe the result.** Tell Spark what to show, which screen to use, and when.
3. **Publish and check.** Ask Spark to publish, inspect a screenshot and playback
   records, then confirm the result on the physical screen yourself.

Spark tasks keep running in Google’s cloud even when your device is off, but the
screenRIG Player is the opposite: it must stay open on the device driving the display,
so keep a browser Player in the foreground while trying it out. A Spark task cannot keep
your screen online on its own.

Already have a Player code? You can combine the request:

> Use screenrig.ai for my lobby screen. My pairing code is [code]. Create a welcome
> board using my events calendar. Ask me for anything else you need.

If the Player never appears in `screenrig screen list`, confirm the Player is open and
its setup is complete, then send the pairing code again.

## Optional sources and schedules

Spark can work from optional connected sources — Google apps and custom connected apps
— and from skills you define. Connections are off by default: approve only the access a
task needs, and keep menu spreadsheets, calendars, or other screen sources to what you
want Spark to read.

Spark schedules control when the agent runs a task; screenRIG playlist schedules
control when content plays on the screen. They are separate. Add a recurring Spark task
only after your first publish is verified, using sources you have already authorized.
The task can refresh and republish content, but your screen’s Player must still be open
to show it.

## If setup stalls

- **Spark cannot run the bundled CLI.** Stop and name the missing capability, then use
  [Claude Code](https://screenrig.ai/docs/start/claude-code/) or
  [ChatGPT Codex](https://screenrig.ai/docs/start/chatgpt/).
- **`version` fails on a Node error.** Ask Spark to check `node --version` and ensure
  Node.js 20.11 or newer is installed in its execution environment.
- **An account command reports `not_enrolled`.** This installation holds no account
  credential yet: new account `screenrig agent enroll --email ADDRESS`, existing account
  only `screenrig agent connect`.
- **`screen list` returns nothing.** Confirm the
  [Player](https://screenrig.ai/docs/players.md) is open, share its pairing code, and ask
  Spark to complete `screenrig screen pair CODE` before listing screens again.

Keep working with the same Spark task rather than creating another account. Full table:
[When the first run fails](https://screenrig.ai/docs/start.md#when-the-first-run-fails).

## Next

Back to [Start](https://screenrig.ai/docs/start.md), open the
[Players guide](https://screenrig.ai/docs/players.md), or
[Compose a still](https://screenrig.ai/docs/compose.md).