It asks first. Until you tell it not to.
Three levels of trust. It starts by drafting, and goes further only when you say so.
Chat
It answers and shows. It touches nothing.
Cowork
It drafts and lines things up. You send, edit or skip.
Agent
It acts inside the limits you set, and tells you after.
Rules it cannot talk itself out of
These are not instructions it was given. They are stops in the code, so nothing it reads in your inbox can argue it past them.
- Nothing more to someone who said no, or asked for later, until you reopen them.
- No third follow-up in a row to someone who has not answered.
- No reply to a conversation it has not just read.
- A draft is held back if they wrote again after it was written.
- A link you never gave it is never sent on its own say-so.
- A daily limit on every kind of action, for each LinkedIn account.
Worth asking before you start.
Does it send anything without asking me?
Not unless you tell it to. It starts by drafting: you send, change a word, or skip. You can let it send on its own for the things you trust it with, and take that back at any time under Account. At every level it asks before it deletes anything, spends credits, posts publicly or changes your profile.
Is my LinkedIn account safe?
Every kind of action has a daily limit for each account, such as 20 invitations and 50 messages a day, and the limits are built in rather than left to a setting. LinkedIn’s own terms still apply to your account, and automated activity can lead LinkedIn to limit one. The limits lower that risk; they do not remove it. HeyReagent is not affiliated with LinkedIn.