Anthropic Unifies Memory Across Claude Chat and Cowork

Anthropic has merged the memory systems behind Claude's chat interface and Claude Cowork, eliminating the need to re-explain context when switching between the two products. The change, announced Tuesday, means information Claude learns in one surface is now automatically available in the other.

Previously, Claude chat and Cowork operated with separate memory, forcing users to restate project details, headcounts, deadlines, or other context every time they moved from planning a task in chat to executing it in Cowork. Anthropic's own example: drafting a conference agenda where Claude already knows the city, speaker list, and headcount from earlier chat conversations, rather than requiring the user to re-enter that information in Cowork.

The mechanics of memory storage are also changing. Claude will now add topics to memory continuously as a conversation happens, rather than generating a summary only after a chat ends. Anthropic says this makes memory updates available faster and lets users move between chat and Cowork even mid-conversation, without waiting for a session to close.

Anthropic is also giving users direct visibility into what Claude has stored. A new interface lets users read, edit, or delete memory entries by topic, rather than relying on Claude's internal summarization with no way to inspect or correct it.

On privacy, Anthropic says Claude will not store sensitive personal information by default — including health data, race, ethnicity, religious beliefs, political views, or gender identity. Users can opt in via an "include sensitive topics in memory" toggle, and the app will notify them whenever a sensitive topic gets saved. Some categories are excluded outright regardless of settings: government-issued ID numbers, Social Security numbers, criminal history, and immigration status, per Anthropic's acceptable use policy.

The unified memory feature is enabled by default for Free, Pro, and Max plan users across web, desktop, and mobile. iOS and Android users need to update to the latest app version to get it. Anthropic did not disclose whether the feature extends to Team or Enterprise plans, or whether API access to Claude is affected.

What this means

This is a product integration change, not a new model release — no new weights, benchmarks, or pricing are involved. The move signals Anthropic's intent to position Claude as a single continuous assistant rather than a set of disconnected products, a structural gap that has let competitors like ChatGPT's persistent memory feel more cohesive by comparison.

The practical value here is real for anyone using Claude for multi-step work: planning in chat and executing in an agentic tool like Cowork is a common workflow, and forcing users to re-brief context at every handoff is a genuine usability failure that this fixes. The bigger open question is trust — giving users read/edit/delete control over memory is the right move, but opt-in sensitive-topic storage means Anthropic is betting users will engage with the toggle correctly, and any mishandling of stored personal data across two merged systems raises the stakes for exactly the kind of privacy scrutiny that memory features in ChatGPT and other assistants have already faced.