Claude gets a memory upgrade

Starting every new AI chat by re-explaining your role, your tech stack, and your company workflow is exhausting.

Anthropic is finally fixing that context fatigue. The company has rolled out a native memory feature for Claude, initially launching it for Team and Enterprise subscribers before expanding access to Pro and Max plan users.

Built for work, walled by projects

Long-term memory is not a brand-new concept in generative AI, but Anthropic built Claude's version specifically around workplace workflows.

The standout mechanism is project-scoped memory. Instead of tossing everything you discuss into one giant, messy memory pool, Claude creates an isolated memory for each project. Your product roadmap specs will not bleed into external client deliverables, and sensitive strategic notes remain separated from everyday operations.

Users also get granular controls. A centralized memory summary in settings shows you every single detail Claude has retained. You can view, edit, or delete memories at any time, or simply instruct Claude in normal conversation to remember or ignore specific preferences. You can even import memory from other AI tools or export your Claude memory for backup.

For conversations where you want zero footprint, Anthropic launched Incognito chats for all users. These sessions do not appear in your chat history and never write to memory. Enterprise admins can also disable the entire memory feature workspace-wide.

Anthropic noted it conducted extensive safety evaluations before expanding the rollout, testing edge cases like whether persistent memory could reinforce harmful behavioral patterns, encourage AI over-accommodation, or be abused to bypass safety guardrails.

Why it matters

Anthropic's approach to memory feels practical because it treats memory like organized workspaces rather than an open personal diary.

For everyday pros and developers, the value is immediate. You no longer have to waste tokens and time pasting system prompts or business context into every fresh tab. Claude simply picks up right where you left off.

Across the competitive landscape, this eliminates one of the biggest sticky advantages OpenAI held with ChatGPT. Many users stayed with ChatGPT simply because it already knew their habits. By adding memory and offering import and export options, Anthropic is actively removing switching friction.

The main friction point, in our view, will be memory hygiene. If Claude latches onto a temporary edge-case preference and treats it as a permanent rule, debugging agent drift across long-running projects could get tricky. Giving users a clear memory dashboard helps, but teams will likely need to treat memory curation like regular project maintenance.