Chat on the desktop

Chat works as it does in the browser: same conversations, same projects, same custom GPTs, same models, synced across every device you sign in on. Chat documents the features themselves. This page covers what is different on the desktop.

Attaching files

Drag files from Explorer or Finder straight onto the conversation, or paste them from the clipboard. The app reads them from disk, so a large PDF does not go through an upload dialog first.

Projects

Projects work exactly as in the browser and are documented under Projects: a project holds related conversations under one shared context, with instructions and files that every chat inside it uses.

On the desktop:

• Projects appear in the sidebar. New project creates one, with a name, an icon and a colour.
• A project's window has two tabs: Configuration for instructions and files, and Conversations for the chats inside it.
• Manage access sits at the top of the Configuration tab. Use it to share the project with colleagues or groups, as described under Share a Project.
• Drop files onto the project window to add them as project files.
• A conversation can be moved into a project, into a different project, or back into the plain history, either from the sidebar row or from the row inside the project.

The sidebar

Below New chat the sidebar holds Special Apps, the gallery of special applications (with + to create one), AI Meetings if meeting recording is enabled for you, and Integrations. Then come your projects and your history.

To act on several conversations at once, hold Shift and click to select a range, or hold Cmd/Ctrl and click to add single conversations. Right-click one of the selected rows, and the menu applies to all of them. Right-clicking a row outside the selection clears it.

Tools in the composer

The Tools button in the composer works as described under Tools: under Capabilities you switch Web Search, Image Generation and Agent Mode, and Special apps below them starts a chat with one of your special applications. Two things are specific to the desktop:

• Image Generation has a submenu where you choose the Image model.
• Cmd/Ctrl+Shift+M switches Agent Mode on and off from the keyboard.

Searching

Cmd/Ctrl+K searches your conversations. It matches titles and, unless your organization has switched that off, the text of the messages, so you can find a chat by something that was said in it rather than by its title.

Diagrams

If a reply contains a Mermaid diagram, it is rendered as an actual diagram rather than shown as code, with a switch in the header between the diagram and its source. This works in Chat, in Code and in Cowork.

Asking for one is enough: "draw me the flow as a diagram".

The window lives in the tray

Closing the window with Cmd/Ctrl+W hides it, it does not quit the app. That is deliberate: the system-wide shortcuts, reminders and scheduled runs only work while the app is running. Quit it properly from the tray icon (Quit) when you really want it gone.

A left click on the tray icon opens the quick input without bringing the whole window forward. Its menu offers Open DeutschlandGPT, Quick input and Quit.

Offline

Conversations you opened recently stay readable without a connection, since the app keeps a local cache. Sending a new message needs the server, and the app says so with a banner instead of failing silently.

If the connection breaks, the banner names the actual reason (DNS, TLS, timeout, HTTP status) and offers a full diagnosis you can copy or send to support. That report is the fastest way to tell a proxy problem from an outage.

Notifications and reminders

Long-running work reports back through the operating system's notifications, and you can ask the assistant for a reminder in any mode. Both are covered on Reminders and notifications, and switched on under Notification settings.

Appearance and zoom

Settings, General holds the light, dark and system choice, the interface language, and a zoom control that scales the whole interface. These are per device, not per account. See General settings.