# Give Claude or ChatGPT long-term memory with MCP

Save notes once in Hjarni. Connect with MCP. Claude and ChatGPT read your notes and remember what you told them, across every conversation.

## Sound familiar?

Built-in memory keeps a few facts. Your notes, research and decisions are not in it.

You re-explain your project, your stack, your preferences. Again. Then again tomorrow.

Your notes live in ten places. Your AI can read none of them.

## How to give Claude or ChatGPT long-term memory

Two steps. No plugins. No scripts.

1. **Save notes once**

   Write in Markdown. Organize with notes, folders, tags. This is your [knowledge base](https://hjarni.com/mcp-knowledge-base).

2. **Connect with MCP**

   Hjarni has a built-in MCP server and is listed in both connector directories. Connect [Claude](https://hjarni.com/docs/connect-claude-mcp) or [ChatGPT](https://hjarni.com/docs/connect-chatgpt-mcp) one time.

One MCP server URL

```
https://hjarni.com/mcp
```

That is the whole setup. Now Claude reads your notes. ChatGPT reads your notes. Ask either one anything and it answers with what you saved. Tell it to write a new note and the note is there next time. Write once, you both remember.

## What "long-term memory" actually means here

This is not a chat-history trick. It is a knowledge base your AI reads live. You decide what goes in: notes, folders, tags. You can read it, edit it, and export it any time. It is plain Markdown you own.

Set instructions per folder. Tell Claude and ChatGPT how to format, what tags to use, what tone to write in. They follow your [instructions](https://hjarni.com/folder-instructions).

## Why a knowledge base, not built-in memory

[ChatGPT memory](https://hjarni.com/compare/hjarni-vs-chatgpt-memory) and Claude's memory both keep a compact profile of you inside one assistant. Neither reads the other's, and neither holds your notes, your research, or what your team decided.

Hjarni is the opposite. One knowledge base, both AIs, in a format you control. Give Claude long-term memory and give ChatGPT long-term memory from the same notes. No copying context between them.

> I use Hjarni for my thesis. Having my assistant search and edit my notes directly has completely replaced the tedious back-and-forth of pasting context into external chat windows.

Wiba

## Simple on purpose

We left features out on purpose. No databases. No kanban boards. No page builders. Just notes, folders, tags, and an MCP server that works. The constraint is the feature.

## Free to start

25 notes free. No credit card required. No trials, no hidden fees.

## How do I give Claude or ChatGPT long-term memory using MCP?

Create a free Hjarni knowledge base. Save a few notes in Markdown. Then connect your AI once: install Hjarni from the Claude Connectors Directory or the ChatGPT Apps Directory, or add the Hjarni MCP server in any other client. From then on both read your notes and remember what you told them between sessions. You can also ask them to write new notes back to your knowledge base.

- [How to give your AI long-term memory](https://hjarni.com/docs/ai-long-term-memory)
- [Knowledge base for ChatGPT and Claude](https://hjarni.com/knowledge-base-for-chatgpt-and-claude)
- [Connect Claude to MCP](https://hjarni.com/docs/connect-claude-mcp)
- [Connect ChatGPT to MCP](https://hjarni.com/docs/connect-chatgpt-mcp)
- [Hjarni vs ChatGPT memory](https://hjarni.com/compare/hjarni-vs-chatgpt-memory)
- [Move your ChatGPT memory to Claude or Gemini](https://hjarni.com/move-chatgpt-memory-to-claude)
- [MCP knowledge base](https://hjarni.com/mcp-knowledge-base)
- [What is MCP?](https://hjarni.com/docs/what-is-mcp)
- [Set up a memory](https://hjarni.com/setup-memory)