🧠 Anthropic Just Upgraded Claude Memory 2.0 WAY More Powerful (Full Breakdown)

Introduction

After using Claude for a while, you may notice one annoying problem. Chat can remember your work and preferences, but Cowork used to miss a lot of that context, so you often had to explain things again.

→ So Claude memory is starting to fix this problem.

Claude Chat and Cowork can now use the same saved information, so you don’t have to repeat the same story every time you switch modes.

→ Shared memory works properly when Cowork runs in the cloud. When you use local mode, Chat and Cowork don’t sync memory in the same way.

In this article, I’ll show you how Claude memory works, how you can turn it on and manage it, and what changes between cloud and local Cowork.

I. What Just Changed with Claude Memory?

Let me give you the quick version first, then we'll go deeper.

Claude memory lets Claude remember useful information about you, then bring that context into future conversations automatically. That memory is shared between Claude Chat and Claude Cowork.

Whatever Claude learns in Chat is available in Cowork, and whatever comes up in Cowork flows back to Chat. It works both ways, and there's no option to keep the 2 separate.

1. What Claude Can Remember

Things like:

• Your project status and deadlines

• Your recurring priorities and workflow preferences

• How specific people (like your manager) prefer updates written

• Your writing tone, formatting style, and technical preferences

So if you've been chatting with Claude for months about a product launch and you now ask Cowork to draft a status update, Claude already has the context. You don't need to brief it again.

2. How Memory Updates Now

Before this update, Claude waited until a conversation ended and then summarized it into memory. That meant a delay, and sometimes things slipped through.

Now Claude adds relevant topics to memory while you're still chatting. You can pause or reset memory anytime from Settings.

Here's the breakdown:

The sensitive topics (health, beliefs, and the like) can be switched on in Memory settings if you want, say, so Claude remembers your gluten allergy when suggesting recipes. But the bottom three rows are never stored, even with sensitive topics on.

The same goes for anything that breaches Anthropic's usage policy, and Claude tells you when it can't save one of these.

3. Who Gets Memory?

• Free, Pro, and Max plans: memory is on by default.

• Team and Enterprise: admins control it, and it's off for individual users until turned on. Heads-up for admins: turning org memory off deletes every user's entries immediately.

• Available on: web, desktop, and mobile. If memory isn't showing on mobile, update your app.

II. How to Set Up and Manage Claude Memory

Good news, there's not much to configure. But start with the time-sensitive bit.

1. Turning It On and Finding Your Memories

Memory should already be on if you're on Free, Pro, or Max. Here's how to check and manage it:

1. Open Claude and go to Settings.

2. Click Memory.

3. You'll see a list of saved topics, which are short files organized by subject.

4. From here you can read, edit, or delete any individual memory.

That last part is genuinely useful. Instead of dumping everything into one long document, Claude stores memories as separate topic files. So if your company changed names, you fix it in one file and every future conversation gets it right from then on.

2. Editing, Deleting, and Cleaning Up

Memory can get old over time. Projects end, preferences change, and Claude occasionally saves something you didn't mean to keep. Here's what you can do:

• Update a memory: just tell Claude the new information in chat, and it updates while you're talking.

• Delete a specific memory: go to Settings → Memory, find the file, and delete it.

• Review sensitive content: check the sensitive-topics toggle if you've turned it on.

• Reset everything: you can wipe memory entirely from Settings for a fresh start.

One thing worth knowing: deleting a chat doesn't delete the memory that came from it. That memory sticks around until you remove it in Settings → Memory.

And if you work in Projects, each project keeps its own separate memory, so context from one won't automatically show up in another.

3. Importing Memory From ChatGPT or Gemini

This one's pretty cool, honestly. If you've spent a long time in ChatGPT or Gemini and you're switching to Claude, you don't have to start from zero.

Anthropic launched a memory import tool at claude.ai/import-memory in early March 2026, available on paid plans (Pro, Max, Team, Enterprise). Here's how it works:

1. Go to Settings → Memory and choose Start Import. On older accounts still on the legacy experience, it's under Settings → Capabilities → Memory.

2. Claude gives you a ready-made prompt. Copy it.

3. Paste that prompt into ChatGPT (or Gemini, Grok, Copilot, and so on).

4. The other assistant generates a summary of your stored context and preferences.

5. Copy that output and paste it back into Claude's memory settings.

6. Claude updates its memory and picks up where you left off.

The whole thing takes about a minute, though it can take up to 24 hours for imported memory to fully activate.

One honest caveat: Claude's support docs note that memory imports are still experimental, so Claude may not incorporate everything perfectly. It's worth reviewing what comes in, because if your old assistant "remembered" something wrong about you, you don't want to carry that mistake into Claude.

III. How Claude Memory Flows Between Chat & Cowork

This is the part that makes the whole thing actually useful, so let me walk through both directions.

Direction 1: Chat Memory to Cowork

Say you've been chatting with Claude about a conference you're planning. You've mentioned the headcount, the city, the speakers, and the budget. All of that is now in memory.

When you go to Cowork and say "draft a conference logistics document," Claude doesn't need you to re-explain any of it. It already knows.

To make sure Cowork is actually using the right memory for a task, and not just guessing, here's a prompt worth running first:

Review what you already remember about my work, writing preferences, and current projects.

Then show me only the memory that is useful for this task: [describe your task].

Tell me which saved details you are using before you start. If something important is missing or old, tell me instead of guessing.

After that, complete the task using the relevant memory and keep the result consistent with my saved preferences.

This way you can see exactly what Cowork is pulling from, instead of hoping it remembered the right things.

Direction 2: Cowork to Chat Memory

It works the other way too. If something new comes up during a Cowork session, like a changed deadline, a new workflow preference, or updated project details, Claude can save it back to memory, and your next Chat conversation already knows.

When you want to save something specific without letting extra task context bleed in, try this:

Update my memory with the information below:

New information: [add the new preference, project update, or work context]

Compare this with what you already remember and replace any old information that conflicts with this update.

Save only the details that may be useful in future conversations. Do not save temporary comments or unrelated context.

After updating the memory, tell me exactly what changed.

After that, hop back into Chat and ask about the new information. If the memory transferred correctly, Chat will already have it.

IV. Cloud Cowork vs. Local Cowork

Okay, so here's the thing I flagged at the top.

Shared memory between Chat and Cowork only works when Cowork is running in the cloud.

If you're using local Cowork, where Cowork runs directly on your machine, it doesn't use the shared memory at all. So you might set everything up perfectly and still feel like Cowork has no idea who you are, and the culprit is simply that you're in local mode.

Here's the difference:

So if Cowork suddenly "forgets" a preference that Chat knows perfectly well, don't panic. Just check which mode you're in. Claude hasn't lost its memory. You're only in the wrong room.

V. Does Claude Memory Work with Claude Code?

Sad to say, but Claude Code is still sitting outside the shared memory system for now.

Right now, memory is shared between Claude Chat and Claude Cowork only. Claude Code keeps its own separate memory workflow and doesn't automatically pull from what you've built in Chat or Cowork.

I know, it's a little odd, since they're all part of the same Claude family. But that's where things stand today. Think of Claude Code as a sibling who hasn't moved into the shared house yet.

→ So for now, treat Claude Code as its own context. If you need it to know something specific about your project or preferences, you'll still tell it directly.

Conclusion

Claude memory got a lot more useful this week. Here's the short version:

The two-way flow between Chat and Cowork is what makes this genuinely useful rather than a nice-to-have. Context you build in conversation is available when you go to do the work, and context from the work comes back to conversation.

Just keep an eye on what's accumulating in there. A smart memory only helps when it's remembering the right things.

If you are interested in other topics and how AI is transforming different aspects of our lives or even in making money using AI with more detailed, step-by-step guidance, you can find our other articles here:

• GPT 6 Could Be Taking Shape as Gemini 4 Gets Ready for the Next Big AI Fight

• Grok Bot Just Killed 1,000 Startups? An Insanely Easy AI Agent Anyone Can Use 24/7

• Want to Sell AI Workflows? Start With These 5 High-Demand Business Automation in 2026*

• This OpenRouter Trick Literally Lets Anyone Run Claude Code for Free or Nearly Free (Full Guide)*

• NEW ChatGPT Work is the Claude Cowork Killer? Step-by-Step Analysis & Full Review*

  *indicates a premium content, if any