Anthropic has introduced a memory feature to its Claude app, designed to help users and teams maintain context across conversations. This update allows Claude to remember details about projects and preferences, reducing the need to repeat information and supporting ongoing complex tasks.

The memory feature is optional, with controls that let users decide what information Claude retains. Alongside this, Anthropic is adding Incognito chats, which do not save to conversation history or memory. These changes are now available to users on Team and Enterprise plans, with Enterprise administrators having the option to turn off memory for their organization. Incognito chats are open to all Claude users.

How Memory Supports Work Activities

Claude's memory is tailored for professional use, focusing on team processes, client requirements, project specifics, and priorities to aid productivity. For instance:

1. Sales teams can retain client details across multiple deals.
2. Product teams can keep track of specifications through development cycles.
3. Executives can monitor initiatives without needing to re-establish context each time.

When users work with projects in Claude, the system assigns a separate memory to each one. This separation helps prevent overlap between different areas, such as keeping product launch plans distinct from client-related work or isolating confidential discussions from routine operations. It acts as a boundary to manage multiple initiatives without mixing details.

Users can access a memory summary in their settings, which compiles all retained information in one view. This summary can be edited directly, or users can chat with Claude to refine what it focuses on or ignores, prompting adjustments to the referenced memories.

Incognito Chats for Flexible Use

For times when users prefer no lasting record, Incognito chats provide a fresh start without drawing from or adding to existing memory. This is suitable for private brainstorming, strategy talks, or any conversation where prior context is not needed. Regular memory and history remain separate and unaffected. For Team and Enterprise users with memory enabled, standard data retention policies continue to apply.

Approach to Implementation

Anthropic notes that the memory feature brings new safety aspects and is built with work environments in mind, steering clear of sensitive conversations. The company is adopting a step-by-step rollout to assess its performance across various user scenarios before wider availability.

Steps to Begin Using Memory

To activate memory, users can enable it in settings and allow Claude to build it from past chats during setup. Once active, queries like "what were we working on last week?" can draw on remembered details from chats and connected tools. Users can also import memory data from another AI tool or export it from Claude for backup or transfer purposes, following provided instructions.

This update supports the idea that productive work develops gradually, with each interaction building on the previous ones through retained context.