How to connect apps to your dot

A dot without connected apps can only work with what you tell it. Connecting your calendar, inbox or workspace tools is what lets it prepare, follow up and notice things on its own. It is also the step with the longest-lasting privacy consequences, because what your dot learns from an app stays with it. This page covers how connections work, what OpenAI says about permissions, and what each on/off switch really does.

How app connections work

Two things follow from that. First, your dot doesn’t only use an app when you ask. It can look through connected information in the background. Second, what it finds can become part of its memory, which you currently can’t edit item by item. See memory and reset.

Permissions are shared with ChatGPT, Work and Codex

This is the detail most people miss. According to OpenAI’s safety FAQ, “plugin permissions are shared across dots, ChatGPT, ChatGPT Work, and Codex.” You manage them in ChatGPT’s Plugins tab, and a change there applies across all four. The help center adds that “the Plugins screen uses shared ChatGPT settings. Available apps may differ.”

How to connect an app

OpenAI’s help center documents this path in the ChatGPT mobile app:

1. Open your dot’s profile.
2. Go to Customize → Plugins.
3. Pick the app and review its permissions before connecting.
4. Connect it. To change or remove access later, use ChatGPT’s Plugins tab.

OpenAI hasn’t published separate desktop steps for connecting apps to a dot. Because the Plugins screen uses shared ChatGPT settings, the Plugins tab in ChatGPT is the place to manage connections on any device. We will add desktop steps after our own setup walkthrough.

Before you connect: a permission checklist

• Would you give this app to ChatGPT and Codex too? The permission is shared, so the answer has to be yes for all four.
• Read the permission list, not just the app name. The mobile flow asks you to review permissions before connecting. Look for anything that lets your dot change or send, not only read.
• Assume it will be remembered. Your dot can form memories from connected apps, and disconnecting won’t remove them. If the app holds something you’d want gone later, the only documented way to delete it is to reset your dot.
• Set your rules before your apps. Custom rules decide whether your dot may send, share or buy through a connection. See custom rules.
• Keep passwords out of chats and plugins. OpenAI’s secure sign-in, which sends credentials to the browser environment “without exposing them to the model,” does not cover passwords you share in a chat, document or plugin.
• Start with one app. Add the next once you’ve seen how your dot uses the first.

What connecting, disconnecting and resetting do

OpenAI’s help articles spread this across several sections. Here it is in one place.

Action	New access to the app	What your dot already learned	The permission in ChatGPT, Work, Codex
Connect (Customize → Plugins, or the Plugins tab)	Starts	Your dot can review the app proactively and form memories	Shared across all four
Disconnect (Plugins tab)	Stops	Not deleted, it stays in your dot’s context	Changes apply across all four
Reset your dot (profile → ••• → Reset)	Not documented	Deleted, along with its conversations and memories	Not documented

Channels: Slack, Teams and email

Some “apps” are also places you talk to your dot.

• Slack. Supported, and set up from desktop. OpenAI says to “use desktop to connect messaging channels, such as Slack or texting.”
• Teams. The announcement says you can message your dot in Slack and Teams. OpenAI hasn’t documented the Teams setup steps in its help center yet.
• Email. You can connect your personal email. “At launch, you cannot give your dot its own standalone email address.”
• Texting. A limited beta for Pro users in the US. See texting your dot.

Your laptop is a separate kind of connection, off by default and set up from the desktop app on that computer. See local computer access.

What your dot can do with connected apps on its own

When your dot researches in the background, OpenAI says it “can read information from permitted connected sources and save private notes,” but its research tools cannot directly send messages to other people, change content through plugins, or control a browser or computer. Custom rules can’t lift that limit.

When your dot does act, another safeguard layer applies. Before it sends an email, for example, OpenAI says Auto-review checks the recipient and the message. For recurring work that reads from connected apps, see scheduled tasks.