August 25, 2026 · TechCrunch · VentureBeat · Anthropic

Claude Cowork starts at one not zero: shared memory changes agent usage

Anthropic is connecting Claude memory between chat and Cowork. The idea is that you do not have to explain a project again when you move from conversation to asking it to act on files or tasks.

• Memory
• Agents
• Productivity
• Privacy

What changes

Before, a user could talk to Claude for days about a project and then have to explain it again when working with Cowork. With shared memory, the agent can reuse context between conversation and action.

The real hook

This focuses on assistants on something more continuous: less copying and pasting context, more delegating a task with known information. For normal users, it may be more important than another model benchmark.

Risk

The same memory that saves work requires clear controls. TechCrunch points out that Anthropic allows reading, editing or deleting what is saved, and there are limits to sensitive information. It's a good news if the user control is visible and easy to use.