Personalize Harvey with User-Level Memory
Learn how to save, apply, and manage personal preferences in Harvey using Memory.
Last updated: Oct 7, 2026
Overview
Memory lets Harvey save your personal preferences and apply them in future sessions.
With Memory, Harvey can remember substantive, reusable information about how you work — such as your role, analytical approach, negotiating posture, or risk priorities — in addition to stylistic preferences.
As you work in Harvey, your preferences are saved and refined. You can also add, edit, and remove memories yourself from your personal settings.
Key Features
- Save preferences from any conversation: Tell Harvey to remember something, or let Harvey recognize an implied preference on its own. Either way, Harvey shows you exactly what it saved.
- Save multiple preferences at once: State several preferences in one prompt, and Harvey saves each one as a separate, individually editable memory.
- Apply memories across sessions and surfaces: Saved memories carry forward into new threads and apply across the web app, workflow and playbook runs, as well as the Harvey add-ins for Word and Outlook.
- Maintain full visibility and control: See memory citations wherever a saved memory shapes a response, and manage all your memories in your personal settings.
Before You Start
Check if Memory is turned on for you in your personal settings. Go to Personal Settings → Memory and check that Use Memory in responses is toggled on. If you don’t see an option for Memory, contact an admin to confirm it’s enabled in your workspace.
Admins can grant the User Memory permission for your role to let you use Memory in Harvey. Refer to our Access Control & Permission Management article for details on managing role permissions.
Getting Started with Memory
The first time you go to Personal Settings → Memory, select Personalize Harvey to be guided through adding your first few preferences. After you’ve answered some questions, Harvey creates your first few memories. You can immediately edit and refine these memories, or manage them from your Memory settings page later.
After Harvey has saved your first memories, select Continue personalizing to open a new thread where you can answer additional questions and save more detailed preferences, or select Settings to go back to your settings menu.
Save and Apply Your Memories
Save a Memory
In a Thread
When working with Harvey in a thread, Harvey can save memories in two ways:
- When you explicitly ask Harvey to remember something (e.g. “Remember to use plain language with no internal shorthand.”)
- When you’ve said something that reflects a stable preference or fact about you (e.g. “I prefer my responses to be concise and start with the bottom line up front.”)
In both cases, Harvey shows you exactly what it memory it has captured in the thread. You can edit or a delete the saved memory by hovering over it and selecting the Edit memory or Delete memory button.
In Your Personal Settings
- Go to Personal Settings → Memory.
- Select Create memory.
- Describe the preference you’d like to save, then select Create.
Apply a Memory
Memories apply automatically after they’ve been saved. Start a new thread and run a relevant query, and Harvey applies your saved memories to the response.
When a memory shapes a response, Harvey attaches a memory citation so you can see which memory was used.
These citations are visible only to you. If you share a thread, collaborators can see the answer but not the memory citations behind it.
Note: If your prompt conflicts with a saved memory, your prompt wins. Harvey treats your current instruction as the most relevant guidance, and may briefly reason through the conflict before following your prompt. To exclude a memory or category of memories from a specific query, just say so in your prompt.
Memories are applied in threads within the Harvey web app, Harvey add-ins for Word and Outlook, and the Harvey Mobile App for iOS. They can also be applied when creating or running playbooks, workflows, and custom agents.
Refer to the table below to see which Harvey actions enable you to save and apply memories:
Action | Saves New Memories? | Applies Existing Memories? |
|---|---|---|
Querying in the Harvey web app | ✅ | ✅ |
Querying in the Harvey for Word add-in | ✅ | ✅ |
Querying in the Harvey for Outlook add-in | ✅ | ✅ |
Querying in the Harvey mobile app | ✅ | ✅ |
Creating a playbook | ✅ | ✅ |
Running a custom agent | ✅ | ✅ |
Querying the LexisNexis knowledge source | ✅ | ❌ |
Running a playbook | ❌ | ✅ |
Running a workflow | ❌ | ✅ |
Creating a workflow in Workflow Builder | ❌ | ✅ |
Creating a custom agent in Agent Builder | ❌ | ✅ |
Using review tables | ❌ | ❌ |
Using Email Harvey | ❌ | ❌ |
Using Harvey for Microsoft Copilot | ❌ | ❌ |
Manage Your Memories
Edit or Delete a Memory
In a Thread
In a thread, you can ask Harvey directly to update or forget existing memories — for example, "Forget everything you know about my NDA preferences."
In Your Personal Settings
- Go to Personal Settings → Memory.
- Hover over an existing memory to show the Edit memory or Delete memory buttons.
- Select Edit memory to modify an existing memory or preference.
- Select Delete memory to remove the memory from Harvey.
Example Memory Use Cases
The table below provides some examples of how you can personalize Harvey with memory.
Category | Description | Example |
|---|---|---|
Voice, Tone & Formatting | Keep outputs in your preferred style, such as word choice, tone, and document layout. | “Always spell out dates in full. Never numeric formats like 9/9/26.” |
Scope, Jurisdiction & Routing | Define what you’re authorized to opine on, and where to route what falls outside that scope. | “Admitted only in New York. Flag anything requiring Delaware or other local counsel.” |
Risk & Analytical Approach | Shape how Harvey reasons: how it frames risk, prioritizes sources, and stress-tests conclusions. | “Pressure-test the answer. Explain what a regulator or board member would attack.” |
Guardrails & Compliance Checks | Proactively flag a risk or verify a condition before finishing a task | “Never let ‘pro forma’ appear without checking whether Reg S-X Article 11 applies.” |
Review, Drafting & Prep Protocols | Define step-by-step process rules for recurring work product, such as redlines, witness prep, or review passes. | “In redline summaries, only flag deviations from the playbook.” |
Recipient-Specific Communication Habits | Adjust delivery based on how one specific, named person reads or reacts. | “The GC reads email only on his phone. 150-word limit, ask in the subject line.” |
Client & Practice Context | Include standing background facts about your role, company, or authority. | “I sign contracts up to $250K TCV; GC signs up to $1M.” |
Security, Data, and Privacy
Memories are text snippets stored within your existing Harvey environment, under the same security posture as the rest of Harvey, and are not used to train models. Each user's memories are private to that user, not shared with other users or pooled across your organization. Admins cannot view or edit anyone’s memories.
Memories persist until you delete them and are not subject to your workspace's standard retention policy.
You can view the Trust Whitepaper on User-Level Memory in the Harvey Trust Center.
FAQs
Was this article helpful?