Le Chat memory: review and reuse useful context

Find the current Knowledge Base controls, understand the move from Memories, and check which saved details should influence future answers.

Updated 5 min read

Le Chat memory is moving from the legacy Memories feature to Knowledge Base in Vibe Work. To find the current controls, use Context → Knowledge → Get Started in the sidebar, following Mistral’s Knowledge Base guide (opens in a new tab). Use it for reusable personal context; keep agreed work and the next step in fenbs when another assistant or person also needs to continue.

Why is the memory menu different now?

Le Chat became Vibe (opens in a new tab), with existing accounts, conversations, and saved content moving with it. Knowledge Base succeeds Memories, and existing Memories migrate into a knowledge topic. Current Knowledge Base works in Work mode, so an old tutorial showing a Chat menu is not a universal setup guide.

Mistral says legacy Chat is being removed progressively, with an Enterprise migration window. If Chat is still available in your account, the legacy route is Chat → Intelligence → Memories → ... → Memories in Vibe. Use the legacy Memories guide (opens in a new tab) only for that remaining interface.

Start by identifying the mode you are actually using. Do not hunt for an old menu indefinitely or assume a missing legacy control means your saved information disappeared. The important check is what the current Knowledge section contains and what you want it to keep.

How do you check what is saved?

  1. Open Work and use Context → Knowledge. If this is your first visit, choose Get Started.
  2. Review the stored entries before asking the assistant to rely on them.
  3. Pick a harmless fact that is still useful, such as your preferred answer length.
  4. Ask Vibe to remember or save that fact, then return to the stored knowledge to check the result.
  5. Check any outdated or overly broad entry before using it for important work.

Knowledge Base is enabled by default unless an organization administrator disables it. The guide does not identify a paid-only restriction or establish identical menus for every phone app. If the feature is unavailable in an organization account, ask about the organization setting before treating it as a personal setup error.

Choose information that stays useful across several conversations. “Use short paragraphs in my drafts” is easier to maintain than a long account of everything you did yesterday. A temporary status belongs with the work it describes, where you can see when it changed.

Does a saved fact appear in every answer?

No. Mistral says retrieval is non-deterministic: the assistant selects stored knowledge when it appears relevant, rather than inserting every saved entry into every response. Saving a fact and seeing it used are different events. A single answer that omits a preference does not by itself establish that the entry was deleted.

Test with a question where the saved detail matters. If you saved a preference for brief answers, ask for a short explanation and inspect the result. If the response ignores an important constraint, state it explicitly in the current request and check the stored entry afterward. Avoid building a critical approval process around silent recall.

For instructions that must be clear in a request, see the Mistral system prompt guide. For rules and agreed decisions that need to follow the work between assistants, fenbs keeps a shared record that connected assistants read when starting.

How do you correct an outdated entry?

Current Knowledge Base text is changed by asking the model, rather than editing the entry directly. Read the existing wording first, then ask for one specific correction. For example: “Change my preferred report format from a long narrative to a one-page summary. Keep the preference for plain language.”

  1. Identify exactly what is wrong: a name, preference, date, or scope.
  2. Ask Vibe to change that information without adding unrelated assumptions.
  3. Review the stored result and make sure the old version is no longer misleading.
  4. Try a small follow-up request that uses the corrected information.

Keep facts and decisions distinct. A note that you are considering a new writing style should not quietly become “this style is approved for every document.” When a person has made a decision, record who decided and what the decision covers. That makes future corrections much easier.

How do you stop memory influencing answers?

Click your organization name at the bottom left, open Privacy, find Personalization, and deactivate Knowledge. Mistral says this preserves the stored entries while stopping them from influencing responses. Disabling Knowledge is therefore different from removing what was saved.

To remove the stored collection, use Delete all knowledge. Read the confirmation carefully before proceeding and make sure you have kept any non-sensitive information you still need. If the goal is only to test whether saved context changes an answer, turning the feature off is a different choice from deleting everything.

Current Knowledge Base has sensitive-data saving off by default, and administrators can prevent users from enabling it. Do not apply the old Memories privacy defaults to the new feature. Review the current guide and your organization’s rules before saving information that should have a limited audience.

What belongs outside personal memory?

Keep source collections, personal preferences, and the state of work separate. Le Chat Libraries hold reusable documents; personal knowledge helps tailor answers. A record of what was checked, what a person approved, and where to resume needs wording that another reader can understand without knowing your chat history.

Before ending a useful session, save the result, its limits, and the next action. If you are choosing where to do the next task, Claude versus Mistral discusses that separate choice. The AI memory hub explains other boundaries worth checking before depending on recall.

Say you finish reviewing a draft and leave two changes for tomorrow. Your preference for short paragraphs may remain useful personal knowledge. The unfinished changes need a separate record: which draft you reviewed, what was checked, and which section comes next. Keeping those details together lets another person understand the task without knowing your personal preferences or having access to the original conversation.

How fenbs helps

Never lose the work you did in an AI session

After a useful AI session, save it in fenbs: the key observations, findings, research and context. Then carry on from exactly where you left off, on any device and in any other AI. The next AI you connect reads it at the start of the session, so you never have to explain the same problem again.

Create a free account

Free for up to 5 people and 50 open tasks · No credit card · Works with Claude, ChatGPT, Cursor and more

How to keep the work with fenbs

Keep enough to carry on from exactly where you left off, on any device and in any other connected AI. A few clear sentences about the work are more useful than copying every personal preference into another place.

  1. Create a free account.
  2. Write what you are trying to finish, what has already been agreed, and the next step.
  3. Connect Le Chat to fenbs through Vibe Work’s connector controls. See Integrations for guidance; adding a custom connector requires an administrator.
  4. Before stopping, save what changed and what still needs a person. Ask the next connected assistant to read that record first.

fenbs does not import your Memories automatically or manage Vibe’s Knowledge settings. If all you need is a personal writing preference in one app, its own memory controls may be enough. Use fenbs when the thing to preserve is the work people and assistants need to share.

Questions people ask.

Will my old Memories move to the new feature?

Mistral says existing Memories migrate into a knowledge topic in Knowledge Base. Review the migrated information rather than assuming it is all still current.

Can I directly type over a Knowledge entry?

The current guide says changes go through the model. Ask for the correction and review the stored result afterward.

Why can I still see entries after turning Knowledge off?

Turning it off preserves stored entries. Use Delete all knowledge if your aim is to remove the collection.

Is the old Chat privacy guidance enough for Work?

No. Knowledge Base has its own settings and sensitive-data controls. Review the current Work documentation and the rules your organization requires.

More in Memory in Claude, ChatGPT, Gemini and Copilot

All Memory in Claude, ChatGPT, Gemini and Copilot articles

Start with one thing.

Free for up to 5 people and 50 open tasks · No credit card · Works with Claude, ChatGPT, Cursor and more

There is nothing to set up first. Write one line and you’ve started.