Manage enhanced workspace memories
Enhanced workspace memories let you save workspace-specific facts and instructions so they stay available in future chats in that workspace. Use them for stable context that should persist, such as client names, audit scope, preferred output format, or rules the assistant should keep in mind.
Workspace memories only apply inside one workspace. For broader workspace setup, see Organizing Work with Workspaces. For longer durable guidance, compare memories with How to use project instructions in workspaces before you decide what to save.
What workspace memories are for
Use workspace memories for short facts or directions that should carry across chat sessions in the same workspace. Each saved memory is added to the AI context for that workspace, so you do not need to repeat the same details every time you start a new conversation.
Good examples include the client name, the framework in scope, how you want controls mapped, or a fixed writing preference. Do not use memories for large notes, changing task context, or one-off requests better handled in the current chat.
Add, edit, or delete a workspace memory
Open the workspace where you want the memory to apply.
Locate the Workspace Memories card (marked with a brain icon) in the workspace view.
Type the fact or instruction in the add field and press Enter or click Add.
The memory saves and appears in the list below.
To update a memory, click the edit icon next to it, modify the text, and save. To remove a memory, click the delete icon. Changes apply to that workspace only.
Keep each memory short and specific. Save one fact or rule per memory so it stays easy to review and update.
Manual memories vs auto-detected memories
Workspace memories can come from two sources:
Manually added: Facts you type yourself in the Workspace Memories card. These are labeled with a manual source badge.
Auto-detected: Facts ISMS Copilot picks up automatically during a conversation, such as a client name, framework scope, or deadline you mention. These are labeled with an auto-detected source badge.
Both types work the same way in chat — they are injected into the AI context for that workspace. The source badge helps you see where each memory came from when you review the list.
Auto-detection is controlled by a separate setting. If you want to stop ISMS Copilot from saving new facts automatically, you can turn off memory detection in memory settings. Turning off detection does not delete memories that are already stored.
What you see in chat when a memory is saved
When ISMS Copilot auto-detects a fact during a conversation, you will see a confirmation message in chat indicating that a new workspace memory was saved. This helps you know when context is being added to your workspace.
If you notice an incorrect or unwanted auto-detected memory, open the Workspace Memories card and delete or edit it as needed.
Ownership and validation
Each workspace memory belongs to the workspace and is tied to the user who created it. The app validates memory entries before saving them.
A single memory can contain up to 500 characters.
Each workspace can store up to 100 memories.
If a memory is too long or the workspace has reached capacity, the app will block the save.
If you cannot save a new memory, shorten the entry or remove older memories you no longer need.
How memories affect the token budget
Saved workspace memories are injected into the AI context for that workspace. More memories means more context is sent with each chat, which uses part of the available token budget.
This usually improves continuity, but too many low-value memories can reduce room for the active conversation. Keep only facts and instructions that are still useful. Move longer durable guidance into How to use project instructions in workspaces if that is a better fit.
Troubleshoot workspace memories
A memory does not persist across chats: Make sure you saved it in the workspace Memories section, not only in the conversation.
A memory is not showing up in responses: Check that you are still in the same workspace and that the memory is still listed there.
You cannot save a memory: Shorten the text if it is over the limit, or delete old memories if the workspace is full.
The AI keeps using outdated context: Edit or delete the old memory so the workspace only keeps current facts.
Next, review Organizing Work with Workspaces to keep each client or project separated, and use project instructions for longer standing guidance that should shape work in that workspace.