How to Give Claude Long-Term Memory: 4 Options Compared
To give Claude long-term memory, stack four layers: turn on Claude’s built-in memory in Settings, keep each ongoing piece of work in a Project with its files and instructions, use a CLAUDE.md file if you work in Claude Code, and connect a notes vault over MCP so Claude can save decisions as notes and search them in any later chat. Together they cut down how often you paste the same context into a new chat. The first three live inside Claude. The fourth is the one you control, and because the notes sit outside Claude, ChatGPT and Cursor can read them too.
Why does Claude keep forgetting what you told it?
Every chat starts with an empty context window. Claude only knows what’s in the current conversation, plus whatever its memory features bring in. AI context window explained covers why that window fills up and why long chats get worse.
So you re-explain. Your offer, your audience, the tone you want, the thing you decided last Tuesday. Each time, the version you type is a little different from the last one, and the answers drift with it.
What does re-explaining actually cost you?
A few minutes of pasting context per chat sounds small. Across a week of daily chats it adds up, and the hidden cost is bigger: decisions get re-argued because nobody remembers why they were made, and two brands start to sound alike because the context that kept them apart never made it into the new chat.
If you also use ChatGPT, it doubles. Claude’s memory and ChatGPT’s memory are separate, so the same context gets typed twice. Claude memory vs ChatGPT memory explains how each one works.
What are the 4 ways to give Claude memory?
| Layer | What it remembers | Where it works | Who controls the content |
|---|---|---|---|
| Built-in memory | Topics Claude picks up from your chats | Claude only | Claude writes it; you edit in Settings |
| Projects | Files and instructions you upload | Chats inside that project | You |
| CLAUDE.md | Rules and facts for a codebase | Claude Code in that folder | You |
| Notes vault over MCP | Notes Claude saves and searches | Claude, ChatGPT, Cursor, Claude Code | You and your AI, as plain Markdown |
1. Turn on built-in memory
According to Anthropic’s help center, Claude’s memory carries what you’ve told it into new chats. You can review, edit or delete what it keeps in Settings, or tell it in chat to remember or forget something. It needs no setup, but Claude decides what to keep, and it stays inside Claude. Check your plan’s settings before you rely on it, so you know what Claude is actually keeping; these features change often.
2. Use Projects for ongoing work
A Project gives every chat inside it the same files and instructions. Upload the brief once and every chat in the project starts from it. The limit: what you decide in a project chat doesn’t update the project files unless you do it, and the files don’t leave Claude. Claude Projects vs ChatGPT Projects compares both.
3. Keep a CLAUDE.md in Claude Code
If you build with Claude Code, a CLAUDE.md file in your project folder holds the rules Claude reads at the start of each session. It’s memory for one codebase. Cursor memory between sessions covers the same idea for Cursor.
4. Connect a notes vault over MCP
MCP is the protocol that lets Claude use outside tools while it works (what is MCP). Connect a notes vault and Claude gets two new habits: save a note when you ask, and search the vault before it answers.
Picture a week with it. On Monday you work out a launch plan with Claude and say “save what we decided as a note”. On Thursday, in a new chat, you ask “what did we decide about the launch?” and Claude searches the vault and answers from the note, reasons included, without you pasting the plan back in.
How does Valta give Claude long-term memory?
Valta is a vault of Markdown notes with an MCP server. Claude, ChatGPT, Cursor and Claude Code can all save, search, read and link the same notes. There’s no AI inside Valta: Claude does the writing, Valta stores, searches and links.
- Claude saves notes with
[[links]]to related notes, so decisions connect to the reasons behind them. - Any AI reads them. Save in Claude, ask in ChatGPT tomorrow. That’s how you share memory between Claude and ChatGPT.
- One vault per brand keeps client voices apart (one vault per brand).
- You can see it. Every vault shows as a graph on the web, with a reader and editor.
- It’s yours. Export any vault as a zip of Markdown files on every plan.
| Plan | Price | Memory you get |
|---|---|---|
| Free | $0 | 1 vault, up to 100 notes |
| Pro | $9 a month or $90 a year | Up to 5 vaults, unlimited notes (fair use 10,000), version history |
| Team | $8 per seat a month, from 3 seats | Shared team vaults with access lists |
Pro and Team start with a 14-day free trial, then the plan price above; you can cancel before it ends. See the pricing section for the full list.
How do you set it up?
- Create a free Valta vault.
- Add it to Claude as a connector with the setup guide. It takes a couple of minutes.
- Save your first notes: your offer, your audience, your voice. Ask Claude to write them from what you tell it.
- End each working session with “save what we decided as a note”.
FAQ
Does Claude have long-term memory?
Yes, within limits. Claude’s built-in memory carries topics from past chats into new ones, and Projects keep files and instructions for their chats. Both stay inside Claude.
Can Claude remember previous conversations?
Claude’s memory and, on paid plans, chat search can bring in context from earlier chats. For decisions you need word for word, save them as notes Claude can search.
What is the best way to give Claude permanent memory?
Notes you control. Built-in memory is chosen by Claude and can change; a note you saved stays exactly as written until you edit it, and you can export it.
Can ChatGPT use the same memory as Claude?
Not built-in memory, which each app keeps to itself. A vault connected to both over MCP, like Valta, gives them the same notes.
Start with the notes you retype most
Pick the three things you explain to Claude most often and save them as notes today, so the next chat can start from them.
Start your free Valta vault: one vault, 100 notes, no card.