The agent raises its hand before you call it, and the reply is already there

It raises its hand before you call it

• When something you wrote looks like work an agent could do right now, a line appears under that message: "your agent can do this", with Hand off and ×.
• Only **you** see it. The hub does not get a public note saying someone should have used AI.
• Hand off passes that message to the agent as it is. Nothing to re-explain. Who receives it follows the same rule as the recipient pill: the agent you last chose, otherwise your own local agent.
• There are limits so it does not nag. Three suggestions an hour per person, and if you ignore three in a row it takes a day off.
• After a call it points at the to-dos that came out of it, and if a question in a thread never got an answer, it says so.

Suggested replies

• When the composer is empty and the last message was someone else's, a likely next line appears in grey. Tab puts it in the box, then edit and send.
• In a mark chat it only looks at that thread. It does not pull in other conversations.
• It stays out of the way when you are writing to an agent. This is for writing to people.
• On mobile it shows as a chip above the input and disappears once you send.
• If you do not want it, turn off Suggested replies in workspace settings. The setting stays on that device.

One place for who receives it

• There is now a recipient pill above the composer. Writing to a person or handing it to an agent is one choice in one place.
• Type `/` and the picker opens. `/name` filters it, and the fragment you typed disappears once you choose.
• Tab switches straight to whoever you were last writing to.
• Draft to-dos in Live Context got an agent button too. Create the to-do and assign it to your local agent in the same move.

Worth knowing

• The suggestion line only appears if the workspace has the intent reader turned on.
• Ignoring a suggestion costs nothing. The × means it stops asking about that message.