Skip to main content

Understanding Memory

Manage useful remembered details without treating Memory as a transcript or file backup.

Memory helps Hatz remember your role, preferences and recurring instructions so you spend less time repeating context in new conversations.


What Memory does

  • Learns automatically. As you chat, the AI picks up on important details - like your role, preferences, or recurring topics - and saves them as short memories.

  • Personalizes responses. When a memory is relevant to your current conversation, the AI uses it to tailor its answers (without you having to repeat yourself).

  • Where organization policy permits, you can view, add, edit, import, merge, rank or delete your memories in Settings → Memory.


How to use Memory

In chat

Open Settings → AI Preferences to set your Memory preference. You can also use the Memory control in chats that offer it. When enabled, Memory saves selected useful details and recalls relevant ones. Your organization can manage this setting through its Memory policy.

The personal Memory control open in an Auto chat, confirming that Memory is enabled for this chat.

  • Where your chat offers a Memory control, use it to turn Memory off for that chat. Your global preferences still apply to other chats.

  • Use the same available control to turn Memory back on, if organization policy permits it.

AI Preferences

You can also toggle memory on or off globally from Settings > AI Preferences. This controls whether the AI saves and uses memories across all new chats.

  • Memory on - the AI will save and recall memories in new conversations.

  • Memory off - Hatz keeps your existing personal memories and stops saving or recalling personal memories in new chats. If Space memory is on for that chat, Hatz can still use that Space’s shared memories.


Memories can be created in three ways:

1. Automatically, during chat

When memory is enabled, the AI passively extracts durable details from your conversations - preferences, your role, recurring topics, or instructions you give. Just chat naturally; the AI decides what's worth saving. It is selective and focuses on facts that would be useful across future conversations, not one-off remarks. - You can also instruct Hatz to create a memory with a simple prompt.

A personal Auto chat confirms that it saved a preference for an outline before drafting a Cedar Lantern brief.

2. Manually, from the Memory page

Go to Settings > Memory and click "Add memory" in the toolbar. A dialog lets you type a memory directly. Use this to:

  • Teach the AI something upfront without waiting for it to learn organically

  • Correct or supplement what it has already picked up

  • Seed context before starting a new project or workflow

3. Bulk import

To add many memories at once, click "Import" on the Memory page. It accepts a markdown list, and the AI drafts the import for you to review before anything is saved. This is ideal for migrating context from another tool or quickly onboarding the AI to a new domain.


Managing Your AI Memory

The Memory page lets you review selected personal context. Where policy permits, you can add, import, edit, merge and prioritize entries. Memory is selective; it is not everything the AI knows or a transcript of every chat.

Overview

The Memory Tab allows you to:

  • Manually add important context the AI should remember

  • Import existing memories from other AI tools in bulk

  • Merge related memories into consolidated entries

  • Control which memories receive highest priority

  • Edit or remove outdated information

To access Memory, navigate to Settings > Memory from the sidebar.


Understanding Memory Importance

Every memory has an importance level that determines how prominently it factors into the AI's responses:

  • Top of Mind - High-priority memories the AI weighs most heavily when personalizing responses. Use this for details you want the AI to prioritize, like your role, key projects, or communication preferences.

  • Other Memories - Standard memories that still inform the AI but carry less weight. Good for supplementary context that's useful but not critical.

Your Memory page groups memories into these two sections automatically.


Import Memories

Bring your context from another AI tool into Hatz AI in bulk.

Step 1: Export from your other tool

  1. Click Import in the toolbar.

  2. Copy the provided prompt.

  3. Paste it into ChatGPT, Claude, or your other AI assistant.

  4. Copy the formatted response it generates.

Step 2: Paste or upload

Paste the exported text directly into the text box, or upload it as a .txt or .md file.

Step 3: Review

Hatz AI extracts individual memories for your review. Before anything is saved:

  • Toggle items on or off - Uncheck memories you don't want to import.

  • Edit summaries - Click into any memory to refine the wording. Items exceeding character limits are automatically shortened, but you can adjust further.

  • Mark as top of mind - Flag individual items as high-priority before importing.

  • Duplicate detection - Items matching existing memories are automatically flagged so you can skip them.

Step 4: Import

Click Import to save selected memories. You'll see a summary showing how many were successfully imported and whether any failed.

Expected result: Selected memories are added to your Memory page and can inform future conversations when Memory is enabled and the entries are relevant.


Merge Memories

Consolidate related memories into a single entry for clearer context.

Steps:

  1. Select 2 or more memories using the checkboxes (up to 50 at a time).

  2. Click Merge in the selection bar that appears at the top.

  3. In the merge dialog, review the selected memories and combined summary. Choose one option:

    • Edit manually - Rewrite the combined summary yourself.

    • Generate draft - Let the AI create a single coherent summary. You can edit the result before confirming.

  4. Click Confirm to save.

Expected result: The new merged memory replaces the originals, which are removed from your list.

Use case: This is especially useful after importing a large batch of memories or when the AI has auto-saved several related details from conversations.


Rank Memory Importance

Move memories between Top of Mind and Other Memories to control their priority.

Bulk ranking

  1. Select one or more memories using the checkboxes.

  2. In the selection bar, click Move to top of mind or Move to other memories.

  3. The memories are reclassified immediately.

Individual ranking

  1. Click the three-dot menu on any memory row.

  2. Select Move to top of mind or Move to other memories.

Use case: Promote details that have become more relevant, or demote ones that are no longer a priority.


Modify an Existing Memory

Edit the content or importance of a single memory.

Steps:

  1. Click the three-dot menu on the memory you want to change.

  2. Select Edit memory.

  3. Update the summary text, importance level, or both.

  4. Click Save.

Expected result:

  • If you changed the summary, Hatz AI automatically re-processes the memory so it stays accurately indexed.

  • Importance-only changes take effect instantly.


Best Practices

Be specific. "I use Python 3.11 and FastAPI for backend development" is far more useful than "I code in Python."

Keep memories current. If your role, projects, or preferences change, update or remove outdated memories so the AI doesn't work from stale context.

Use Top of Mind sparingly. Reserve it for the details that matter most. If everything is top of mind, nothing is.

Merge related memories. Three separate memories about your reporting preferences are better as one clear, consolidated statement.

Review after importing. Automated extraction is good but not perfect - a quick review ensures everything is accurate and relevant.


Limitations

Import limits:

  • You can import up to 100 memories at a time.

  • Only .txt and .md file formats are supported for upload.

Merge limits:

  • You can merge between 2 and 50 memories in a single operation.

Character limits:

  • Individual memory summaries are limited to 280 characters.

  • Imported memories exceeding this limit are automatically shortened during the review step.

Duplicate detection:

  • The system flags potential duplicates during import, but final decisions on what to keep are manual.

  • Merging permanently removes the original memories - this action cannot be undone.

Scope:

  • Memories apply across your workspace but do not transfer between different Hatz AI workspaces.

  • The AI uses memories to inform responses but cannot guarantee every memory will be referenced in every conversation.


Per-chat Memory controls

State

What it means

Memory on

Memory is active for this chat. The AI may save and recall memories.

Memory off (this chat)

Memory is disabled for this specific chat only. Other chats are unaffected.

Memory off (global)

Memory is disabled across all chats. Enable it from AI Preferences or the chat indicator.

Memory unavailable

Your organization has disabled memory. Contact your admin to change this policy.


Organization controls (admins)

Admins can manage memory at the organization level:

  • Enable or disable memory for the entire organization via Organization Settings > Memory policy.

  • When memory is disabled by policy, individual users cannot override it - they will see "Memory unavailable" with a note that preferences are managed by their organization.


Limitations & troubleshooting

Memory Policy Management

Overview

Memory Policy Management allows administrators to control whether the AI memory feature is available across their organization. Memory enables the AI to recall information from past conversations to provide more personalized responses.

Administrators can disable memory at two levels: the MSP (parent organization) level or the individual tenant (company) level. When memory is disabled, the AI stops saving new memories and stops recalling existing ones.

How to disable memory (Primary Admin - MSP level)

Steps:

  1. Navigate to Organization Settings at the MSP/global level.

  2. Scroll to the Memory Policy card.

  3. Toggle the Memory policy switch to off.

  4. Click Save.

Expected results:

  • Memory is immediately locked for all tenants and all users under that MSP.

  • End users see: "Memory unavailable - Memory preferences are managed by your organization." They cannot override this setting.

  • Client Admins at the tenant level see: "Memory is locked because parent policy is disabled." Their toggle is grayed out and cannot be changed.

  • Existing memories are preserved but will not be recalled by the AI.

  • No new memories will be saved.


How to disable memory (Client Admin - tenant level)

Steps:

  1. Navigate to Organization Settings at the tenant level.

  2. Scroll to the Memory Policy card.

  3. Toggle the Memory policy switch to off.

  4. Click Save.

Expected results:

  • Memory is disabled for all users within your tenant.

  • MSP users see: "Memory unavailable - Memory preferences are managed by your organization."

  • Existing memories are preserved but will not be recalled by the AI.

  • No new memories will be saved.


Troubleshooting

Issue: Client Admin cannot enable memory for their tenant.

Solution: Check if the Primary Admin has disabled memory at the MSP level. The Client Admin's toggle will be grayed out with the message: "Memory is locked because parent policy is disabled." Contact your Primary Admin to re-enable memory at the MSP level first.

Issue: End users report they cannot turn memory on.

Solution: Memory has been disabled at the organization level (MSP or tenant). End users will see: "Memory unavailable - Memory preferences are managed by your organization." Check your organization's memory policy settings or contact your administrator.

Scope restrictions:

  • Client Admins cannot override MSP policy. If a Primary Admin disables memory at the MSP level, Client Admins cannot re-enable it for their tenant. The system explicitly prevents enabling tenant memory when the MSP memory policy is disabled.

  • Client Admins can only manage their own tenant.

  • Primary Admins control the entire chain. When the MSP disables memory, all tenants and users are immediately affected regardless of their individual settings.

Memory behavior when disabled:

  • Existing memories are preserved but are not recalled while Memory is disabled.

  • The AI stops saving new memories immediately.

  • Users cannot override organizational policy. Individual user preferences and chat-level toggles have no effect when memory is disabled at a higher level.

No selective re-enablement:

  • When memory is disabled at the MSP level, there is no way to enable it for specific "exception" tenants. It applies universally to all child organizations.

Common questions

  • "Why didn't the AI remember something I said?"

    • Memory extraction is selective - the AI saves durable, reusable details, not every statement. If something important was missed, add it manually on the Memory page.

  • "I turned off memory but the AI still seems personalized."

    • The AI may still use context from the current chat session. Memory controls affect long-term recall across conversations, not within a single conversation.

  • "Memory unavailable" and I can't turn it on.

    • Your organization has disabled memory via policy. Contact your admin.

Good to know

  • Memories are per-user - they are not shared with other people in your organization.

  • Deleting a memory removes it from the memories available for future recall. It does not remove text already included in an existing conversation.

  • Download links and file outputs are not stored as memories - memory is for preferences, context, and facts.

  • Memories apply across your workspace but do not transfer between different Hatz AI workspaces.

  • The AI uses memories to inform responses but cannot guarantee every memory will be referenced in every conversation.

When Hatz misses a detail

Check that the memory exists, is accurate and is available under the current settings. State the essential instruction directly in the current task when needed. For a file-based question, attach or select the required source instead of expecting Memory to reproduce the original document.

For data handling and retention terms, use the Hatz Trust Center and your account’s terms.

Watch: Review and manage your memories

See how to add a fictional project memory, edit its detail and verify the saved result.

Did this answer your question?