On March 2, 2026, Anthropic made Claude's memory feature free for all users — a strategic move that coincided with the QuitGPT boycott of ChatGPT and made switching from ChatGPT to Claude significantly easier. Memory is one of the most impactful and least understood features in any AI platform. When it works well, it transforms Claude from a stateless tool that forgets everything after each conversation into a persistent assistant that knows who you are, how you work, and what you care about. When it works poorly — or when users do not know how to control it — it creates a creeping sense that the AI knows too much, or that it is carrying forward incorrect assumptions from old conversations. This guide covers everything you need to know to use Claude memory well.
What Claude Memory Actually Does — The Plain English Explanation
Claude memory stores specific facts, preferences, and context from your conversations and applies them automatically in future sessions. Without memory, every conversation starts completely blank — Claude does not know your name, your profession, your ongoing projects, or any preferences you have expressed before. With memory enabled, Claude retains the things it learns about you across sessions, so you do not need to re-explain your context every time you start a new conversation.
- What Claude remembers: Your name and how you prefer to be addressed. Your profession or primary use case. Ongoing projects you have mentioned by name. Communication preferences (formal vs. casual, brief vs. detailed). Technical context (programming languages you use, subjects you study, tools in your workflow). Explicit preferences you have stated ('always give me bullet points,' 'skip the preamble,' 'use metric units').
- What Claude does NOT remember by default: The specific content of past conversations — it remembers facts extracted from them, not the full conversation text. Sensitive information unless you explicitly ask it to store something. Memories from before memory was enabled.
- How storage works: Claude generates memory entries automatically from conversations, or users can create them explicitly. Memory entries are short factual statements ('User is a software engineer working in Python,' 'User prefers concise responses without extensive caveats'). You can view, edit, and delete every memory entry at any time.
- Privacy: Memory data is stored separately from conversation history. Deleting a conversation does not delete memories extracted from it. Deleting a memory entry removes it immediately and it is no longer applied to future conversations.
How to Import Your ChatGPT Memory Into Claude
Anthropic launched a memory import feature specifically designed for users switching from ChatGPT. The process transfers your stored ChatGPT memory items into Claude, so you do not lose the context you have built up over months of ChatGPT use.
- Step 1 — Export from ChatGPT: Open ChatGPT → Settings → Personalization → Memory → View Memory. Screenshot or copy the memory items you want to transfer. Alternatively, go to Settings → Data Controls → Export Data to download a full data export including your memory.
- Step 2 — Tell Claude directly: In a new Claude conversation, paste your ChatGPT memory items and say: 'Please store these as memories so you remember this context going forward.' Claude will confirm each memory stored.
- Step 3 — Verify and refine: Go to Claude Settings → Memory to review what was stored. Edit any entries that need correction and delete anything you do not want Claude to remember.
- Step 4 — Add what ChatGPT missed: ChatGPT memory items are often generic. Use this migration as an opportunity to add more specific context — your current projects, your communication style preferences, and technical context about your work.
How to Control and Manage Claude Memory
- View all memories: Claude.ai Settings → Memory (or Claude app → Profile → Memory). Every stored memory entry is visible, editable, and deletable. Review these monthly — Claude sometimes extracts memories that are imprecise or no longer accurate.
- Delete individual memories: Click the X next to any memory entry to remove it permanently. Claude will stop applying that context immediately. No conversation history is affected.
- Pause memory for a conversation: Start any message with 'This is a private conversation — please do not remember anything from this session.' Claude will not extract memories from that specific conversation.
- Create explicit memories: Tell Claude directly what to remember: 'Remember that I prefer all code examples in Python 3.11' or 'Remember that I am studying for the UPSC Civil Services exam in 2026.' Claude will confirm storage.
- Clear all memories: Settings → Memory → Clear All. This resets Claude to a blank memory state. Useful if you want to start fresh or if stored memories have become outdated or inaccurate.
- Memory and shared accounts: If you share a Claude account with others, memories from your conversations will apply to all users of that account. Use separate accounts or pause memory when sharing.
The 8 Memory Configurations That Make Claude Dramatically More Useful
- Your professional context: 'I am a [role] working in [industry]. My primary tasks involve [specific activities].' This single memory entry improves relevance across every conversation.
- Your communication style: 'I prefer direct responses without unnecessary caveats. Skip phrases like "certainly" and "great question." Be concise.' This eliminates the most common AI response patterns that users find annoying.
- Your technical stack: 'My primary programming language is Python 3.11. I use FastAPI for web services, PostgreSQL for databases, and deploy on AWS.' Claude will default to these technologies in all code examples.
- Your ongoing projects: 'I am building [project name], a [description]. The tech stack is [stack]. Current status: [status].' Claude will have context for project-specific questions without re-explaining every time.
- Your study context: 'I am preparing for [exam] in [month/year]. My weak subjects are [subjects]. I prefer explanations using analogies before technical definitions.' Transforms Claude into a personalised tutor.
- Your location and unit preferences: 'I am based in India. Use metric units, INR for currency, and Indian date formats.' Eliminates constant clarification about units and context.
- Response format preferences: 'When I ask for code, always include: a brief explanation before the code, inline comments for non-obvious logic, and a usage example after the code.' Consistent output format without repeated instructions.
- Your learning style: 'Explain concepts by starting with a real-world analogy, then the technical definition, then an example. Do not use jargon without defining it first.' Every explanation Claude gives will be structured for how you actually learn.
Pro Tip: The most overlooked memory configuration: telling Claude how you want it to handle uncertainty. Try: 'When you are not sure about something, say so explicitly and tell me your confidence level. Do not hedge everything — only flag genuine uncertainty. When information might have changed since your training, tell me to verify it.' This single memory entry dramatically improves the quality of Claude's responses by calibrating it to signal when you should double-check versus when you can trust its output.