Do My Companions Share What They Know About Me? Yes - the Memory Is Yours, Not Theirs
By The Fellowi Team · · 5 min read

A question that arrives in support every week, in some form: "I told Lena I was starting a new job, and then Marcus asked how the new job was going. Are they talking to each other?" They are not. They are reading the same notebook, and the notebook is yours. This post explains what that means precisely, because "the AI remembers you" is a phrase that hides three different mechanisms.
One memory list per account
When a conversation contains something durable about you - a name, a city, a job, a preference you stated plainly - it may be extracted into a memory. That memory is stored against your account, not against the companion who heard it. When any companion replies to you, the relevant memories from that one list are pulled into the prompt, whichever companion originally learned them. The mechanism itself is described in how companion memory works; the point here is the scope. It is per person, not per character.
So yes: tell one companion you moved to Lisbon, and the next one you open may well ask about the move. That is not a leak between characters. It is the same reason a new companion does not need the questionnaire filled in again.
What stays inside one chat
Two things do not travel. The first is the conversation itself: the messages, the mood, the thread you were following at midnight. The second is the rolling summary each chat keeps of its own recent turns, which is how a companion picks up a scene where you left it. Both are attached to the conversation, not to your account. A companion you have never spoken to knows the facts about you and nothing about the evening you spent with someone else.
The same is true of a group chat: it has its own thread and its own summary, and its participants draw on the same account-level memories.
Where each memory came from
Every memory records which companion it was learned from and which message it came from. Retrieval ignores the first: a fact is a fact, whoever heard it. The second matters in one case. If you delete a companion you created yourself, the memories learned in its chats go with it. Deleting messages, on the other hand, does not delete the memories they produced - a durable memory outlives the sentence it was extracted from. If you want a memory gone, the memory page is the place, not the chat.
The psychologists are the exception
The eight AI psychologists do not extract memories from your sessions at all; that is a data-minimisation rule, not an oversight. What they keep is small and consented: a goal you set, a homework task, a note about what helps. Those few notes are what lets a psychologist ask how the homework went next week. They sit in the same memory list as everything else, labelled, where you can read and delete them like any other entry - the detail is in what AI psychologists remember.
The one switch
Settings has a memory page. It shows every saved memory as plain text, lets you delete one or all of them, and has a single switch at the top. The switch is account-wide: turned off, no companion uses saved memories, and none are retrieved until you turn it back on. There is no per-companion setting, and we would rather say that than imply a control that does not exist. Deleting a memory takes effect on the next message; it is not a request that waits.
One more surface reads the same list: when a companion writes to you first, the message is drawn from what it knows about you, which is this list and nothing else.
Why we built it this way
Because the alternative - a separate memory per character - would mean telling every new companion your name again, and would make the memory page eleven pages. The cost is the surprise this post opened with. If that surprise is unwelcome, the switch and the delete button are yours; and if a companion ever remembers something wrong, correcting it in chat and deleting the old entry is the whole fix. What happens to all of this if you leave is in what your chat data is. And if you would like a companion who starts with a blank notebook, that is what the switch is for.
Questions this post answers
If I tell one companion something, will another companion know it?
Yes, if it was saved as a durable memory. Memories belong to your account, and every companion you chat with can draw on the same list. What is not shared is the flow of a particular conversation: each chat keeps its own running summary.
Can I stop companions from sharing memories with each other?
Not selectively. The memory switch in Settings is one switch for the whole account: on, every companion uses your saved memories; off, none of them do. You can delete individual memories or all of them at any time.
Do the AI psychologists share what I tell them with other companions?
The psychologists deliberately do not extract memories from your sessions. The only things they save are a handful of consented notes such as a goal or a homework task, and those appear in the same memory list you can read and delete.
Keep reading

The Fellowi Lobby: One Anonymous Room for Every Subscriber
A permanent nickname, no names or emails, one-tap translation across 11 languages, and a few house rules. Meet the lobby.

Why a Friendly Companion Steers You to the Adult Category
You asked, and she kindly pointed you somewhere else. That is by design: two kinds of companion, the cards under her reply, and the lines no companion crosses.

A Kindroid Alternative for People Who Would Rather Not Build
One hands you a workbench, the other a room of people already talking. Voice, price, memory and languages, compared with the dates on.