How to edit Claude's memory

Claude has been able to remember past conversations since the summer of 2025. That memory feature, which Anthropic enables by default for all users including free accounts, lets the chatbot personalize responses without repeated context. A recent update also lets the AI carry memories across both regular chats and Claude Cowork sessions. The new editing tools build on that system, and Anthropic has been gradually introducing them over recent weeks.

On the web client, users can start by opening the profile page menu, selecting Settings, then clicking Memory. Under the Topics section, they can press the edit icon to modify a saved memory or the delete icon to remove one entirely. When editing, a prompt bar lets them describe what Claude should change. On mobile, users tap the profile icon, go to Capabilities under the App section, and select Memory files. After picking a topic, they can use the prompt bar to edit or hit the trash icon to delete the memory.

By default, Claude does not save memories related to sensitive subjects such as health, race, ethnicity, religious beliefs, politics or gender identity. A new option called "Include sensitive topics in memory" can be enabled to allow the recording of such information. Each time Claude generates a memory connected to a sensitive subject, a notice appears. With an older version or the mobile app, users will not see this notice, and Claude will not save the sensitive topic. Some information, such as government ID numbers, is never stored, even with the sensitive-topics option on. The chatbot also does not retroactively save past conversations.

Users who want to stop automatic memory generation can toggle off "Generate memory from chats." This gives two options: pause memory or reset it entirely. Anthropic warns that resetting cannot be undone. After re-enabling the feature, Claude starts from scratch and loses all previous memory.

In projects, Claude keeps memory in a separate silo so that the context window is not overwhelmed by irrelevant data. Users can shape this memory by adding comments, documents and information from GitHub through the right side of the interface. At the end of each day, Claude generates a summary of the work done, which the user can edit. If a conversation is accidentally started outside a project, the three-dot menu provides an "Add to project" option to sort it correctly. To make Claude remember specific items across every tool, users can add them in the Profile section of the Memory menu.

Incognito mode offers a temporary way to avoid any memory writing. Anthropic explains that incognito chats are not saved, added to memory, or used to train models. In addition, a March update added an import tool for moving conversations from other chatbots to Claude. Users can start this process by selecting "Import memory from other AI providers" in the settings.