New Claude memory lets users pull up old conversations on demand

Anthropic’s Claude chatbot can now remember past conversations, but only when users specifically request it to do so.

The company rolled out the memory feature on Monday to Claude Max, Team, and Enterprise subscribers. Users must enable the function in settings and explicitly ask Claude to reference previous chats.

Claude will only retrieve and reference your past chats when you ask it to, and it’s not building a user profile, said Anthropic spokesperson Ryan Donegan to The Verge.

The feature works differently from that of ChatGPT, which automatically stores all conversations unless users opt out. Google’s Gemini goes further by using both chat history and Google Search data to personalize responses.

Claude’s approach requires users to prompt the artificial intelligence (AI) when they want it to search past conversations. The system then summarizes relevant discussions and offers to continue previous projects.

Anthropic demonstrated the feature in a video showing a user asking what they discussed before vacation. Claude searched through chat history, provided a summary, and suggested continuing their work.

The memory function operates across web, desktop, and mobile apps. It keeps different workspaces and projects separate, allowing users to organize conversations by topic or purpose.

To activate the feature, users go to Profile, then Settings, and toggle on “Search and reference chats.” The company plans to expand access to other subscription plans soon.

This launch comes during intense competition between Anthropic and OpenAI. Both companies have been adding features like voice interaction and larger context windows while seeking major funding rounds.

OpenAI released GPT-5 last week. Meanwhile, Anthropic is reportedly pursuing funding that could value the company at $170 billion.

Memory features represent a key strategy for AI companies trying to increase user loyalty. The goal is to make chatbots more useful over time so people stick with one platform instead of switching between services.

However, the feature has sparked controversy. Some ChatGPT users appreciate the convenience, while others worry about privacy or feel uncomfortable when AI recalls personal details.

Claude’s search-based system addresses these concerns by making memory activation deliberate rather than automatic. Users maintain control over when their chat history gets referenced.

The feature focuses on helping users resume interrupted projects without re-explaining context. This proves especially useful for ongoing work tasks or complex research sessions that span multiple conversations.

Privacy advocates may welcome Claude’s approach as it avoids building persistent user profiles. The system only accesses past chats when requested, then returns to its default state.

Whether Claude’s opt-in memory will prove as effective as always-on systems remains to be seen. The company is betting that user control outweighs the convenience of automatic recall.