Kasimir Docs DEEN To the app →

Memory

Kasimir remembers durable facts from your chats – as pages in your notebook, visible, editable and switchable off at any time.

Kasimir remembers durable facts from your chats: which systems you use, what your role is, how you like your answers, what you are currently working on. In every later chat Kasimir already knows these facts – you don't have to repeat them.

Everything remembered is stored as regular pages in your Knowledge Notebook. You see every entry, can change or delete it, and can switch memory off at any time.

Memory is off by default. It must first be enabled for your company (see Switching off) – only then does your personal switch have any effect.

What Kasimir remembers

After an answer Kasimir briefly looks at the conversation and asks itself: Would this sentence still be useful in three months, in a completely different conversation? Only what passes this test is remembered. The task of the current chat, intermediate results or the text you are drafting right now don't qualify – which is why most chats yield nothing.

What is remembered is sorted onto sub-pages under “Memory”. Four of them are standard:

Page

What goes there

Example

About me

Durable facts about you that you state yourself: company, role, field, systems in use, language.

“We use Jira for all customer tickets.”

How I work

How Kasimir should work for you – corrections and preferences about format, length, tone and approach. The entry notes in brackets when to apply it.

“Always give me a short summary first, then the details.”

Ongoing projects

Undertakings that run beyond this chat, or permanent constraints. Each undertaking gets its own page.

“We are migrating to the new ERP by the end of the year.”

Systems & references

Recurring pointers: systems, folders, ticket systems, addresses.

“The quote templates live in the folder Sales/Templates.”

Kasimir chooses the sub-page itself. If you create your own sub-pages under “Memory” – say “Customers” – Kasimir uses them as soon as an entry fits the topic. If a new topic matches no existing page, Kasimir creates a new sub-page with a fitting title and links it in the “Memory” index. You can also tell it in the chat where an entry belongs: “Remember on the sub-page ‘Fleet’: …”

Entries are stored in the language you write in. What is already remembered is not stored twice; if something has changed, the existing entry is replaced. At most three new entries are created per answer.

What Kasimir does not remember

  • Nothing you did not write yourself. Only your own, self-typed sentences count as evidence. Content from attachments, documents, web pages, tool results or pasted third-party text (emails, reports) never becomes a memory – not even when Kasimir summarises it in its answer.

  • No assumptions Kasimir makes about you. What the AI itself claims about you is not evidence.

  • Nothing about third parties – unless you explicitly ask for it (see below).

  • No special categories of data: health, HR, salary or job-application data are never stored.

  • No judgements about you and no confidential figures from third-party text (prices, contracts, terms).

Short interjections such as “yes”, “go on” or “again” do not trigger a memory pass.

Just tell it

You don't have to wait for Kasimir to notice something on its own. Simply write in the chat what it should remember:

  • “Remember that our invoices always go to accounting@…”

  • “Please remember: we don't deploy on Fridays.”

  • “Keep this in mind: Ms Meier is our contact at the supplier.”

  • “Note this down: …” / “Make a note of this: …”

  • In German: “Merk dir …”, “Notier dir …”, “Behalt im Hinterkopf …”

With such an explicit request Kasimir stores exactly that statement – even if it concerns other people, such as contacts or colleagues. You then carry the responsibility just as when typing a note. Health, job-application and salary data still stay out.

The request only applies to the message it appears in. Kasimir does not extend it to earlier statements in the same chat.

What you see and how to intervene

The notice below the answer

When Kasimir has remembered something, a small chip with a bookmark icon appears below the answer: Remembered: … with a short version of the entry. Clicking it opens the notebook page the entry was written to.

Forget

Next to the chip you find Forget. One click removes the entry from the page again and the notice disappears. Kasimir remembers this decision: the same statement will not be remembered again automatically from the rest of the conversation. It only comes back if you later explicitly ask for it again.

The pages in the notebook

In the Knowledge view you find the page Memory under Personal. It is marked “Always on” and contains only a short index: one line per sub-page with a link to it. The actual entries are list items on the sub-pages – the standard pages About me, How I work, Systems & references, one page per undertaking, plus any further sub-pages that you or Kasimir have created for topics of their own. Kasimir only creates its pages once there is a first matching entry.

The pages behave like any other page in the notebook:

  • Edit: rephrase, extend or strike entries – the next answer uses the current state.

  • Rename: the titles are just initial suggestions. Kasimir finds its pages again even after renaming.

  • Delete: a deleted page goes to the trash as usual. When a new entry comes along later, Kasimir creates a fresh page.

  • Not movable: memory pages stay in their area. Moving them to other areas or under other pages is not possible.

Before every new entry Kasimir reads what is already on the sub-pages – including items you wrote or rephrased by hand. That is why it never creates duplicates, and when something has changed it replaces the existing item instead of adding a second one next to it.

The index page is included in every one of your chats; the sub-pages are found automatically when a question matches their content – just like all pages in the notebook (see Knowledge & the AI).

The index is deliberately capped at 20 lines so it doesn't take up too much space with every question. Once all lines are used, Kasimir creates no further sub-pages and files new entries onto the existing pages instead – new undertakings onto “Ongoing projects”.

Projects

If you write in a project chat, what is remembered does not go into your personal memory but onto a separate Memory page in the project's notebook. All project members can see these pages, and they apply in the project's chats.

In projects too, the source is only your own messages – never the contributions of other members. Personal memory and project memory stay separate; project internals do not travel into your private chats and vice versa.

When the project owner deletes the project, Kasimir asks what should happen to the project knowledge including memory: delete it permanently or take it over into the owner's personal notebook. Taken-over memory pages become regular pages there.

Switching off

Two switches determine whether memory runs – both must be on:

  • For you: under Settings → Memory switch off Memory active. Kasimir then remembers nothing new, and your existing memory pages no longer flow into answers either. The pages themselves stay in the notebook until you delete them.

  • For the company: an administrator enables or disables memory under Administration → Chat settings with Automatic memory. It is off by default. If it is not enabled for your company, the settings page shows a corresponding note.

Privacy

  • Processed in a German data centre. The memory pass runs on Kasimir's own models – regardless of which model you chose for the chat.

  • Only your own words. Attachments, documents, web content and statements by the AI never become memories. Nothing is stored about third parties unless you explicitly ask for it; special categories of data are always excluded.

  • Visible and correctable. Every entry is plain text on a page you can edit and delete. Personal memory pages are seen only by you and by the AI in your own chats; project pages by the members of the project.

  • Deleting a chat. If you delete a chat, the link between entry and chat is removed. The entry itself stays on the page until you delete it there – so nothing you wanted to keep gets lost.

  • Off means off. When memory is switched off, no new entries are created and no existing pages are used in answers.