Control a TV with ChatGPT: digital signage from ChatGPT dots or Codex

A chat reply on its own sets nothing up, because the plugin has to run somewhere. Two OpenAI agents run it:

You create a dot in the ChatGPT desktop app or in ChatGPT on a desktop browser. OpenAI's own Get started with your dot covers availability. Your ChatGPT login is separate from your screenRIG account, and your agent never needs your screenRIG password.

1. Send your dot this message:

   Install screenrig.ai using the official plugin at https://github.com/screenrig/plugin and help me set up my first screen.

2. Approve what it asks for, then ask it to confirm the plugin is ready. Your dot works on its own cloud computer, so there is nothing to install on yours.
3. Give it your email. It creates your screenRIG project and emails you an invitation to the dashboard. Nothing waits on that email.
4. Open the Player on the TV and read the code to your dot. Install the Player from the Downloads page, or open the web Player on any screen with a modern browser. Keep it in the foreground while you try it.
5. Say what to show. The dot publishes, asks the screen for a screenshot and tells you what it saw.

Already have the code? Say it all at once: "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."

Run these in your terminal, not inside Codex, then start Codex:

codex plugin marketplace add https://github.com/screenrig/plugin.git --ref main --json
codex plugin add screenrig@screenrig --json

Codex needs Node.js 22.23.4+, 24.22.1+ or 26.12.1+ for the plugin's command-line tool. Run codex plugin list --json to confirm the plugin is registered, and restart Codex if it was already running. The rest is the same as for a dot: your email, the Player's code, then what to show. ChatGPT Codex setup has the full path.

A dot that stays on can own a screen. Ask it once, and it keeps the screen current:

The meetings here are made up. When you do say the word, this is what a recorded session on a test screen looked like. The school and its hallway screens are made up too.

The all-clear was on the screen four seconds after the word that ended the drill. Schools has the whole run, and AI digital signage lists ten more requests to try.

• Publish. Posters, menus, photos, video, web pages, live streams and the 200+ ready-made apps, on one screen or many.
• Schedule. Breakfast, lunch and dinner menus, or a closing notice, each in the screen's own time.
• Take over. One request puts a notice on every screen with a tag, for a set time up to 7 days or until you clear it.
• Check. A screenshot of one screen or of every tagged screen, and the Player's own report that it is playing.
• Watch. A screen offline for 60 seconds raises an event, and a dot that stays on can look into it.
• Keep a routine. Ask for a screen that changes when its data does. A dot keeps working between conversations, so the routine runs without your computer.

• The Player must be open on the device that drives the screen. A dot can publish all day, and the screen plays it only while the Player runs.
• A chat reply alone does not install the plugin. If your dot cannot install or run it, ask what access or capability is missing, or use Codex.
• Review is yours to turn on. With Require review for agent changes, or Propose access, each change waits for your approval in the dashboard.
• Availability and plans are OpenAI's. Follow OpenAI's guide for what your ChatGPT account includes.

Claude uses Claude Code in the terminal or the Claude app, Gemini uses Spark, and Grok uses Grok Bot or Grok Build. How can my AI agent put content on a TV? covers them all, and Developers shows the connected-app route for ChatGPT, the API and the command line.

Can ChatGPT control a TV?

Yes, through a ChatGPT dot or Codex. Both run the official plugin, which pairs the screen, publishes content and checks a screenshot of the screen. An ordinary ChatGPT reply, with nothing running the plugin, controls nothing.

What is the difference between a dot and Codex?

A dot lives in ChatGPT, has its own cloud computer and keeps working between conversations, so it suits a screen that must stay current. Codex runs in your terminal on your machine, so it suits a person who already works there. They install the same plugin and publish the same way.

Do I need a terminal for ChatGPT dots?

No. You send your dot a message and it installs the plugin on its own cloud computer.

Can ChatGPT see what is on the TV?

It can ask the screen for a screenshot and look at it. Publishing also waits for the Player to report the new playlist, up to two minutes, and says whether it is playing.

Can ChatGPT run several screens?

Yes. Tag the screens, and one request changes or takes over every screen with the tag. The answer reports each screen's own result.

Does ChatGPT need my screenRIG password?

No. Your agent signs in with its own credentials. Never paste passwords, invitation links or tokens into the conversation.

Can ChatGPT read my calendar or sheet for the screen?

It reads them with the access you give it, and only what is on the screen goes to the TV. Identify the source and approve only the access you want it to have.