How to Turn Off Claude Memory

To turn off Claude memory for your account, open Settings > Memory, switch off memory generation and choose Pause memory if you want to preserve the existing entries. For just one conversation, start a new chat and turn Memory off in the “+” menu before sending your first message.

Decide whether you want less personalization, a temporary break or permanent removal before choosing a control. If you still want selected work context, fenbs can hold an explicit project brief separately from Claude’s automatic memory.

How do you pause memory for your account?

1. Confirm that you are in the intended account and workspace, especially if you have both personal and work access.
2. Open Settings > Memory. If memory appears under Settings > Capabilities instead, use the legacy section of the official guide.
3. Turn off memory generation and choose Pause memory to stop using and creating memory while keeping existing entries. Read the confirmation before selecting a different option.
4. Check “Search and reference chats” separately if you also want to prevent past-chat searches. Memory and chat search are different controls.
5. Start a fresh conversation for the work you want to do next. Note which controls you changed so you can deliberately review them before turning anything back on.

These account controls are documented in Claude’s memory guide↗ (opens in a new tab). Use the instructions matching your screen rather than assuming a screenshot from another account represents your current settings.

Should you choose pause or reset?

Pause is the reversible choice when you want a break from memory. Reset is the permanent deletion option, including project memories. Do not select reset just to find out whether a less personalized conversation feels better; decide first whether you actually want the saved record removed.

For example, a designer might want to brainstorm an unfamiliar visual style without established preferences steering the first draft. That is a temporary working condition. A change of employer might instead prompt a review of obsolete background. The desired outcome should determine the control you use.

Write a short note for yourself about why you paused it. When you later resume normal work, you can decide whether the old entries are still appropriate. A deliberate return is more useful than turning every setting back on simply because you cannot remember which one changed the answer.

If your goal is selective correction rather than stopping the feature, Claude memory management covers that separate workflow. Keeping those goals distinct prevents a temporary experiment from becoming an unnecessary cleanup.

How do you keep just one conversation out of memory?

Start a new chat, open the “+” menu and turn Memory off before the first message. You cannot change that chat’s setting after it starts. This is useful when you want a one-off discussion without changing the account’s normal behavior.

If you also want the conversation excluded from chat history, use the separate incognito option↗ (opens in a new tab) outside a project. A memory-off ordinary chat remains in history and can be found by searches from other conversations. Choose by what you want retained, not merely by the absence of personalization in the current answer.

Read the mode indicator before entering information. Do not assume that changing browser tabs, using another device or opening a private browser window changes the Claude account’s memory setting. The app’s own control is the relevant check for the conversation you are about to start.

Why might an answer still seem personalized?

Look at the information you deliberately supplied. An attached biography, an instruction naming your preferred audience or earlier messages in the open conversation can explain a tailored answer without proving memory was used. Starting with a fresh, short prompt makes that distinction easier to investigate.

Check profile preferences too. The incognito documentation says these can still apply, so an answer following your preferred writing style is not by itself proof that the memory setting failed. List the context sources before drawing a conclusion from the tone of a single response.

Project sources deserve a separate look: Claude’s project documentation↗ (opens in a new tab) describes uploaded knowledge, instructions and “Use account memory.” Inspect those controls on the project you use, especially if the information you wanted excluded appears in an uploaded brief.

The broader distinction is covered in Claude project memory, while what Claude is explains the product’s place among assistants. For a controlled work brief, fenbs’s context workflow lets you choose which project facts the connected assistant should read.

How can you check the setting without exposing private facts?

Use a harmless preference you already know is saved. In a fresh chat, ask an open question about that preference without stating its answer. Compare what happened with the visible account or conversation setting. Keep in mind that a correct guess is not evidence of a particular storage mechanism.

If the result surprises you, reduce the experiment. Remove attachments, avoid a project for the initial test and do not quote earlier messages. Record whether chat search appeared or a previous conversation was cited. Specific observations help you distinguish a setting problem from context you supplied yourself.

Turning memory off is not a statement that all retained information has been deleted. Review the applicable vendor controls and your organization’s requirements if deletion is the actual goal. This guide concerns future use of memory; it is not a substitute for a records-retention decision.

How to do this with fenbs

1. Create a free account, free for one person with no credit card, and add the project you want to continue.
2. Write a short context note containing only the verified work facts you intend the assistant to use. Keep unrelated personal background out of it.
3. Add tasks for the unfinished work and record continuing instructions in Decisions and rules, with the person who decided them.
4. Connect Claude through Integrations using browser sign-in. MCP, a standard way for an AI app to use an outside tool, lets the approved connection read the project.
5. Ask the assistant to read project context before starting and save progress before stopping. Review the next-step record when you return, regardless of your personal memory setting.

fenbs does not control Claude’s memory settings. If you only want a one-off conversation with no remembered background, the built-in conversation control is enough. For the coding product’s separate memory behavior, see Claude Code memory; other account-level questions are in the AI memory hub.