On Monday, August 12, 2025, Anthropic quietly pushed out a much-requested convenience to its Claude users: the ability for the assistant to search and reference past conversations so you don’t have to paste the same context into every new chat. The company showed the feature in a short YouTube demo where a user asks Claude what they’d been working on before a vacation; Claude flips through previous threads, summarizes the relevant bits and asks if the user wants to resume the project.

This isn’t intended to be a spooky, always-listening memory. Anthropic frames it as an on-demand search of your chat history: turn the feature on in Settings → Profile under “Search and reference chats,” then ask Claude to fetch earlier exchanges when you need continuity across sessions. The rollout started with paid tiers — Max, Team and Enterprise — with other plans to follow. The feature works across web, desktop and mobile and keeps different projects and workspaces separate so your “work” chats don’t get mixed up with your grocery lists.

Anthropic is careful to distinguish this from the kind of persistent, profile-building memory other companies have trialed. According to Anthropic’s spokespeople and reporting, Claude will only retrieve and reference past chats when you explicitly ask it to; it isn’t building a background user profile that constantly shapes responses. That’s a privacy-forward framing meant to give users the productivity gains of continuity without the sense that the bot is silently cataloguing them.

Think of it as a “searchable backlog.” You can tell Claude, “Continue where we left off on the marketing deck,” and it will scan your past chats, pull up the relevant thread, summarize the previous work, and then ask whether you want to keep editing that document or start an offshoot. For teams, that can be especially handy: project threads, research notes and brainstorming fragments that were scattered across days or teammates become accessible without manual copy-pasting.

The tradeoffs

This is where nuance matters. On paper, on-demand recall is a neat middle ground: you get continuity without a standing profile. But it also puts a lot of trust in the search and relevance system — if Claude pulls up the wrong thread or misses crucial context, the continuity becomes noisy rather than helpful. There are also psychological and policy concerns bubbling up: people can become attached to chatbots, and memory features magnify those dynamics; regulators and privacy advocates will probably want to monitor how clearly controls are presented and how easy it is to erase or limit recall.

Anthropic’s move is a predictable next step in an arms race between AI assistants to own more of your workflow. For power users who juggle multi-session projects, being able to ask an assistant to fetch past notes is a real productivity win. For casual users the feature will probably feel optional — something you enable when it helps and disable when it doesn’t. For businesses evaluating assistants, the feature is another checkbox in the “collaboration and data control” column: does an AI let teams continue work easily and keep sensitive data discoverable only when necessary?

Quick practical tips before you try it

• If you’re on Max, Team or Enterprise, look for Settings → Profile → Search and reference chats and toggle it on.
• Treat it like a search tool: use clear project names and consistent labels in chats if you want better retrieval.
• Be mindful of what you store in chats — treat Claude threads like editable notes rather than a private vault.
• If privacy is a concern, test how easy it is to remove a thread or opt the feature off entirely.

Anthropic’s update is a pragmatic, user-control-focused attempt to solve a real pain point: losing context between chat sessions. It stops short of the deeper, always-on profiling some other systems offer — and that positioning may appeal to users who want continuity without a shadow profile. Whether on-demand recall becomes the default user expectation or a niche convenience will depend on how well Claude’s search works in the messy real world of half-finished projects and scattered notes. For now, it’s a simply useful step toward making chatbots less like single-session helpers and more like ongoing collaborators.