Claude adds memory for work teams
Claude is adding optional memory for Team and Enterprise users, with a separate space for each project and controls to review or delete what it remembers. It is also introducing incognito chats that do not save conversations to history or memory, and later expanding the feature to Pro and Max plans.
Anthropic has added memory to Claude so teams can continue their work without re-explaining the context in every conversation. The feature remembers projects, processes, clients, and preferences, while leaving control in the hands of users and company administrators.
Separate memory for each project
Memory is available to users on the Team and Enterprise plans. If you use projects, Claude maintains a separate memory for each one. That keeps information about a launch campaign from being mixed with information about a client or an internal conversation.
In practice, a sales team can preserve an account's context across multiple meetings. A product team can return to a feature's specifications weeks later. And an executive can track the status of several initiatives without reconstructing all the background each time.
Claude brings what it remembers together in a summary that you can review and edit. In Settings, you can check that information, ask Claude to ignore specific data, or adjust which context it should take into account. Memory is optional.
Conversations without memory
Anthropic has also created incognito chats. They let you talk to Claude without adding information to memory or keeping the conversation in your history.
This can be useful for a sensitive brainstorming session, a confidential strategy, or simply starting from scratch. Your regular memory and conversation history remain unchanged. On the Team and Enterprise plans, the organization's data retention policies also apply.
Enterprise administrators can disable memory for their entire company at any time. The feature was initially designed for work, with limits intended to prevent especially sensitive conversations from becoming permanent context.
What changes for you
Until now, using Claude for long-running projects meant repeating instructions, goals, and important details. With memory, the assistant can answer a question like “what were we working on last week?” and use information from previous conversations and connected tools.
That saves time, but it also makes it more important to know what data the tool retains. The ability to review and edit the summary, keep memory separate by project, and use incognito chats are the main controls for preventing unwanted mixing or memories.
Anthropic says it tested the feature in sensitive situations, including conversations related to well-being, to check whether memory could reinforce harmful patterns, lead to overly compliant responses, or make it easier to bypass its safety measures. The company says it adjusted Claude's behavior based on those tests.
The feature initially rolled out to Team and Enterprise plans and was later expanded to Pro and Max. In all cases, it must be enabled from Settings. What matters now is whether memory actually reduces repetitive work without turning accumulated context into a source of errors, leaks, or confusion between projects.