Claude’s new memory import tool is essentially a fast‑track for switching AI assistants without feeling like you’re starting a relationship from scratch.

Why Anthropic built a “memory import” tool

If you’ve been using ChatGPT, Gemini, Copilot, or any other chatbot for months, that system quietly accumulates a lot of context about you: who you are, how you work, what you like, and the projects you keep coming back to. The pain point Anthropic is going after is simple: people want to try Claude, but they don’t want to re‑teach another AI all of that from day one. So instead of building some complex direct integration with every rival, Anthropic did something much more pragmatic: it turned the entire “migration wizard” into a single, shareable prompt and a paste box.

At a high level, the workflow looks like this:

• Claude gives you a prewritten prompt designed to ask your old AI to dump “everything it knows about you” in one shot.
• You run that prompt in your previous AI, get a big block of text describing your memories and context.
• You paste that block into Claude’s memory settings, and Claude parses it into structured long‑term memory that it can actually use.

From your perspective, it feels like dragging your old profile into a new app, but done with copy‑paste instead of an export file.

Step‑by‑step: how the import flow works

The user journey is deliberately short, and Anthropic markets it with the line “Import what matters in under a minute.”

1. Open the import flow in Claude
   • On the web or Claude Desktop, go to Settings → Capabilities → Memory and hit “Start import,” or click the “Import memory” entry point Anthropic is now highlighting on claude.com/import-memory.
   • You’ll see a dialog with a special prompt already written for you, plus an empty text box where you’ll later paste the exported data.
2. Copy the prewritten prompt for your old AI
   • The import dialog shows a long, carefully engineered instruction that tells your current AI: “List every memory you have stored about me, plus any context you’ve learned from past conversations, in one clean block so I can copy it.”
   • You click “Copy” once; that gives you everything you need to take to ChatGPT, Gemini, Copilot, or any other model you’ve been using.
3. Run that prompt in your old AI
   • In a new chat with your previous AI, you paste that prompt and send it.
   • The model then generates a consolidated “memory dump” of what it knows about you: personal basics (name, role, location), recurring projects, tools and tech stack, stylistic preferences, and any other long‑running context it has stored.
   • Different platforms handle “memory” differently, but the prompt is designed to coax out as much structured information as possible in one go.
4. Copy the exported memory block
   • Once your old AI finishes the response, you select the entire output and copy it.
   • Some power users take this chance to quickly prune obvious junk or outdated items before moving on.
5. Paste into Claude’s memory settings
   • Back in Claude, you return to the import dialog in Settings → Capabilities → Memory and paste the block into the input box.
   • You click “Add to memory,” and Claude then runs an internal pass over that text, extracting individual memory entries rather than just storing one giant blob.
6. Review and tweak what Claude learned
   • After the import, Claude breaks that wall of text into discrete “memory edits” you can inspect under “View and edit your memory.”
   • You’ll see items like “User is a project manager,” “Prefers brief and direct communication,” or “Utilizes task management apps, time tracking tools, and collaboration software regularly,” each as its own line you can delete, correct, or extend.
7. Test it in a fresh chat
   • Anthropic even suggests a starter line: “I updated my memory. What did you learn about me?” to let Claude summarize its new understanding of you.
   • From then on, that context is available across new conversations, so your “first” chat on Claude can already reference your long‑term projects and preferences.

The entire process is intentionally built so that a non‑technical user can complete it with two copy‑paste actions and no file handling.

What exactly gets imported?

Anthropic’s help docs and surrounding coverage make it clear that this isn’t about slurping raw transcripts of every past conversation; it’s about importing the summarized profile your old AI has built up about you.

Typical categories include:

• Personal basics: your name, rough location, job title or role, sometimes team or company.
• Work and tools: your tech stack, content or coding tools, frameworks, and platforms you use regularly.
• Ongoing projects and goals: long‑running initiatives, article series, coding projects, or business tasks you frequently discuss.
• Preferences and style: tone you like, formatting habits, things you want the AI to avoid, and how detailed you want answers to be.
• Constraints and rules of engagement: budget ranges, privacy sensitivities, preferred depth of explanation, industry focus.

Claude doesn’t keep this as an opaque black box. Its memory is explicitly positioned as transparent and editable: you can view, edit, and delete individual entries at any time from Settings → Capabilities → Memory. That’s a deliberate contrast to systems where memory quietly accumulates in the background and can be hard to inspect.

Who can use it: free vs. paid

This is where the story has evolved quickly.

• When Anthropic first rolled out memory, it was gated to paid plans (Pro, Max, Team, Enterprise), and import/export was framed as a premium capability.
• More recently, Anthropic extended memory and the import tool to free Claude users as well, specifically to lower the friction for people switching from ChatGPT and other rivals.

That means:

• You can now create a free Claude account and still bring your existing context over via the import flow.
• Paid tiers still offer other upsides — more capacity, faster models, enterprise features — but the basic ability to keep your “AI identity” when you move is no longer paywalled.

There’s also broad support across surfaces: the memory import interface is available on Claude’s web app and on Claude Desktop, not just a single client.

Why this matters for heavy AI users

From a user‑experience perspective, the import tool solves three big problems that power users have been bumping into for the past year.

1. The “cold start” problem
   • Without import, every new AI feels clueless for the first dozen chats; you constantly repeat who you are and what you’re working on.
   • With import, Claude can greet you already knowing you’re a project manager who organizes tasks, uses AI for resource planning, and prioritizes team collaboration.
2. Platform lock‑in
   • When your preferences and workflows live inside a single AI’s memory, that context acts like glue, keeping you there, even if you’d rather try something else.
   • By making it trivial to port that context, Anthropic is effectively attacking lock‑in: your long‑term “AI profile” becomes portable between platforms, not trapped.
3. Trust and control over your profile
   • Because Claude exposes your memory as a list you can read and edit, you can prune outdated roles, dead projects, or personal details you no longer want baked into responses.
   • Experts are already advising users not to just blindly dump years of AI memories into Claude, but to treat this as an opportunity to clean house: keep current tools and projects, drop the noise.

For someone who lives inside AI tools all day, that combination — portability, transparency, and a quick setup — is a subtle but important shift.

Practical tips before you hit “add to memory.”

A few habits emerging from early adopters are worth borrowing if you want this to actually improve your Claude experience, not just clutter it.

• Use the import as a reset, not a dump
  Before pasting into Claude, skim the exported text from your old AI and cut anything obviously stale: previous jobs, old side‑projects, expired subscriptions, or relationships you never want referenced again.
• Keep it focused on how you work
  The most valuable memories often revolve around various types of productivity methods (e.g., “I use time blocking to allocate specific hours for focused work, I follow the Pomodoro Technique for bursts of productivity, and I employ the Kanban system to visualize my tasks”). These approaches tend to be more meaningful than just remembering random facts.
• Check Claude’s interpretation
  After import, open “View and edit your memory” and verify how Claude split things up. Delete any misinterpretations or over‑generalizations before they start influencing responses.
• Test with a real task
  Instead of asking, “What can you do for me?” integrate Claude into your project workflow by saying, “Help me outline the major milestones for my upcoming product launch.”

Under the hood, the memory import tool is “just a prompt and a text field.” In practice, for people who rely on AI as a daily collaborator, it’s a way to take years of invisible training and carry it with you when you decide to switch.