Anthropic gives Claude a unified memory across chat and Cowork

Anthropic just solved one of the more annoying friction points in AI-assisted workflows. Claude now shares a single memory layer between its standard chat interface and the Claude Cowork agentic tool, meaning users no longer have to re-explain their projects, preferences, and working context every time they switch between the two.

What the update actually changes

Before this update, Claude maintained separate memory pools for its chat and Cowork environments. Chat captured things like user preferences and conversational context. Cowork stored project-scoped information. The unified memory now operates in the cloud, syncing context across both surfaces. If you tell Claude in a chat session that your team uses TypeScript and prefers functional programming patterns, Cowork will know that too when you spin up an agentic task.

Anthropic also introduced editable “Topics” within the memory settings. Users can navigate to Settings, then Memory, and see exactly what Claude has stored about them, organized by topic. From there, entries can be edited or deleted entirely.

Privacy guardrails are baked in as well. Sensitive categories, including health information and personal beliefs, are excluded from memory storage by default. Users can opt in if they want.

The rollout isn’t uniform across all plans. Memory is enabled by default for Free, Pro, and Max subscribers. For Team and Enterprise plans, it’s switched off by default.

Building on the Cowork foundation

Claude Cowork launched in January 2026, initially available only to Mac users. It has since expanded to other platforms, evolving from a standalone agentic workspace into something closer to an integrated productivity layer. The tool was designed for knowledge workers who need Claude to do more than answer questions: think multi-step task execution, project management, and autonomous workflows.

But early feedback apparently surfaced a consistent complaint. Users found it frustrating that Cowork didn’t carry over the context they’d built up in regular chat sessions. Every new Cowork session felt like starting from scratch, even if the user had spent hours establishing preferences and project details in the chat interface.

Competitive implications in the AI productivity race

Anthropic’s approach here is notable for its emphasis on user control. Rather than simply making memory more powerful behind the scenes, the company is making it transparent and editable. The Topics system gives users a dashboard view of what Claude knows about them.

The decision to keep memory off by default for Team and Enterprise plans also signals something about Anthropic’s go-to-market strategy. Enterprise customers care deeply about data governance, and letting IT administrators control memory settings rather than forcing them on is likely to reduce friction in procurement conversations.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.