Your AI assistant remembers your job, your projects, sometimes your health. One question is enough to make it tell you: see what it knows, sort it, and leave with a memory that stays on your side.
Try it nowIt asks it to sum up what it has kept about you, section by section.
Analysing the answer requires JavaScript. You can also paste it into « Memory & Context » once in the chat.
Open the chatAfter a few months of use, an AI assistant often knows more about you than you think. Your job, your company's name, the project keeping you busy, the city you live in, sometimes an allergy or a treatment mentioned in passing. These details do not vanish when the conversation ends: most consumer assistants now keep a memory from one chat to the next, on top of the history.
This page does two things. First, it lets you see what your assistant has kept, in black and white. Then it lets you bring over what is useful to you, so you can start on IA Confidential with an assistant that already knows you, instead of starting from scratch.
1. One question to ask. The tool gives you a ready-made question. You ask it to the assistant you already use: ChatGPT, Claude, Gemini or another one. It asks the assistant to sum up what it knows about you, section by section: job, areas of expertise, health, where you live, what you like, what you reject, what has kept you busy these last months. It forbids it to invent anything: a section it knows nothing about must stay empty.
2. Its answer to paste. You copy the answer and paste it into the tool. It is analysed and sorted into the same sections as the « Memory & Context » window of the chat.
3. A check before anything is saved. You read each section again, fix a sentence, remove an item in one click. Nothing is kept until you confirm. Once confirmed, the memory is stored in your browser, and the assistant uses it from your very first question.
An AI that knows you gives better answers. It matches the level of detail to your expertise, takes your country into account for a legal or tax question, avoids suggesting a dish you cannot eat, remembers the current project without you explaining it again.
The downside is that this memory is often kept by the assistant's provider, on its servers. Seeing what it holds is a good habit, even if you do not switch tools: many people find a personal detail they never thought they had shared.
Bringing this memory over means keeping the upside without the downside: the context follows you, but it stays on your side.
The answer you paste is analysed by an open model hosted on our servers in Switzerland, just long enough to sort it into sections, then forgotten: we keep no copy of it.
The memory you confirm is stored in your browser, not on our servers. You can find it at any time under « Memory & Context » to edit or erase it. It is used by the confidential models and is never sent to external models.
If you later create an account from the same browser, it comes along.
No. It only asks it to sum up what it knows. To erase that memory on its side, use its own settings: most assistants let you review and delete memories one by one, or all at once.
No. The tool works without signing up, and so does the chat. A free account then lets you go further, without losing what you prepared here.
It will answer with what it finds in the current conversation, or leave the sections empty. You can also fill everything in yourself under « Memory & Context », once in the chat.
It is your call. An allergy or a treatment matters so that an answer never puts you at risk: the Health section is there for that. If you would rather not keep them, remove them before confirming. To go further on what is best never shared with a consumer AI, read our guide on what not to paste into an AI.
