What's new in Spaces

Chat now offers suggestions at both ends of the thread. Before you start, starter prompts help you open a useful conversation in your Space. After each answer, follow-up suggestions appear, shaped by what you just discussed, so the next question is one click away.

A guided create flow and short onboarding through Chat, Knowledge, and Core Prompt, plus invites when you add people. Where configured, Agent mode supports longer, multi-step work without leaving your Space.

Use '@' to focus your questions on a specific document, and '#' to reference another Space where enabled. Deep search reaches further across your Knowledge when the answer isn't in one place alone, still cited, still inside your perimeter.

A clearer Sources experience: richer file support, document details, and processing status, so you always know which sources are ready to ground answers in Chat.

Pin and rename the conversations you'll return to, delete what you don't need, or start a temporary chat that never lands in History. Your archive stays intentional, not noisy.

Create and switch between Spaces so each team, project, or domain keeps its own Knowledge, Core Prompt, and History. No bleed across Spaces: context stays where it belongs.

Dictate questions when typing gets in the way. Share a conversation with a private or expiring link when someone needs the thread, not access to the whole Space.

Private. Local. Yours. Spaces is a workspace where Chat answers from your Knowledge, shaped by your Core Prompt, with source citations and conversations that are never used to train any model.