Sharing chats

Last published : Nov 09, 2026
Sharing a chat lets people in your project follow it live and take part. This page explains what that gives them, what it doesn't, and why the defaults are set the way they are.

What sharing means

A shared chat still runs in your workspace. When someone you shared it with sends a message, the assistant answers it the way it answers you. It uses your files, your credentials, your connectors, and your model allowance.
That is the point of sharing, and also its risk. A colleague can ask the assistant to change a document without a copy of your workspace. They can also ask it to do anything it can do for you. That's why the Share chat dialog tells you to share only with people you would trust at your keyboard.

Who can open a shared chat

Two things are needed to open a shared chat:
  • Being in the audience you chose: everyone in the project, or the members you selected. One audience applies to all your shared chats.
  • Having that chat's link.
The platform checks the first. You control the second, by deciding who gets the link. A person in your audience can't find your other shared chats without their links.

What people see

A person who opens the link is asked for their name. Everyone in the chat sees who wrote each message.
They see the chat's name, your name, and the conversation as it happens. They don't see your other chats, your settings, or anything in your workspace outside the chat's folder.
You see who opened the chat, and who is there now.

What people can do

By default, people in a shared chat can read it and send messages.
Work with files is off until you turn it on, because it gives away more control. With it on, they can open, download, edit, and add files in the chat's folder, and attach files to their messages. They can't delete, rename, or move files. Hidden files are never shown to them.
Files can't be shared from a chat whose folder is your home folder, or contains it. That folder holds your credentials.

Limits on messages

Every message a guest sends uses your model allowance, possibly while you aren't watching. So each shared chat has an allowance of its own: 10 messages, then one more every 30 seconds. When it runs out, people see how long to wait.