Skip to main content
Assistant memory is a persistent, privacy-safe continuity layer that enables Glean Assistant to remember your preferences, work context, and ongoing projects. Memory helps the Assistant provide more relevant, personalized responses without requiring you to repeat information across sessions.

How memory works

Glean Assistant uses two types of memory to personalize your experience:
  • Saved memories: Explicit, user-controlled information you tell the Assistant to remember (for example, “Remember that I prefer concise answers” or “I’m a Sales Engineer in EMEA”).
  • Extracted memories: Implicit insights mined from your chat sessions and work activity, within retention guardrails. These include your response preferences, role and responsibilities, active projects, and recent topics.
Memory improves response quality and efficiency for tasks that benefit from personalization, including tone, style, role awareness, and continuity across conversations.
Memory does not grant the Assistant access to documents you cannot access. Memory only stores user-level facts and preferences.

Memory categories

The Assistant organizes memories into the following categories:
CategoryDescription
Role and responsibilitiesA summary of your role, job function, and key areas of focus
Response preferencesYour preferred tone (concise vs. detailed), formatting (bullets, tables), formality level, and style
Active projectsCurrent projects and initiatives you’re working on
Recent topicsTopics and focus areas from your recent work activity
OtherAdditional context and facts that don’t fit other categories

Benefits of memory

Memory helps you work more efficiently by:
  • Reducing repetition: You don’t need to re-explain your role, preferences, or project context in each conversation.
  • Providing continuity: The Assistant remembers context from previous discussions, so you can continue where you left off.
  • Personalizing responses: Responses match your preferred communication style and level of detail.
  • Improving relevance: Recommendations and suggestions are tailored to your work focus areas and past requests.

Manage your memories

You can view and delete your saved memories from Settings, and you can view, add, and edit memories directly in chat.

View memories in Settings

  1. Click your profile icon and select Settings.
  2. Select Personalization.
  3. Review your memories organized by category.

Add memories using chat

Tell the Assistant what to remember using natural language:
  • “Remember that I prefer comprehensive responses in tabular format”
  • “Remember that I work on the enterprise sales team”
  • “Remember that I always want code examples in Python”
The Assistant confirms when it saves a memory.

View memories using chat

Ask the Assistant what it remembers about you:
  • “What do you remember about me?”
  • “Show me my saved memories”

Delete memories

You can delete individual memories or all memories: Delete a specific memory:
  • In Settings > Personalization, click the delete icon next to the memory you want to remove.
  • In chat, tell the Assistant: “Forget that I prefer responses in bulleted format”
Delete all memories:
  • In Settings > Personalization, click Delete all to remove all memories in a category.
Deleted memories may be regenerated from your activity data during the weekly memory update cycle.

Runtime controls in chat

You can control memory usage during a conversation using natural language:
  • “Don’t use memory in this chat”
  • “Use memory for this response”
  • “What memories are you using right now?”

Privacy and security

Memory is designed with privacy and security as core principles:
  • User-level data: Memories are stored at the individual user level. There is no cross-user memory sharing.
  • No document access: Memory stores facts and preferences, not document content. Memory does not grant access to documents you don’t have permission to view.
  • Retention policies: Extracted memories are persisted for 30 days and regenerated weekly. Saved memories persist until you delete them.
  • Sensitive data: The system does not extract sensitive personal attributes. Extraction is scoped to work context and preferences.
  • Admin controls: Your organization’s admin can configure memory settings, including enabling or disabling memory features at the organization level.

Known limitations

  • Possible duplication: You may see duplicate entries in memory views. These are deduplicated during the weekly update cycle.
  • Deleted memories may reappear: Extracted memories may be regenerated from activity data after deletion.
  • Fast Mode: Memories cannot be created or referenced when using Fast Mode.
  • Memory edits in UI: User memory edits through chat commands are supported; edits directly in the Settings UI are view-only except for deletion.
  • Weekly updates: Extracted memories are regenerated weekly. Changes may not appear immediately.

Frequently asked questions

Chat history stores your previous conversations so you can reference them later. Memory extracts and stores durable facts, preferences, and context that persist across all conversations, not just within a single chat thread.
Yes, agents can access your memories to provide personalized responses, subject to the agent’s configuration and your memory settings.
Your admin can configure memory settings for the organization but cannot view individual user memories. Memories are private to each user.