Your AI forgets you every time you switch apps. Here’s why, and how to fix it.

Where ChatGPT, Claude, Gemini and coding agents keep what they know about you, why none of it syncs, and three ways to make your preferences follow you.

If you use more than one AI assistant, you’ve noticed that each one knows a different version of you. ChatGPT remembers your diet, Claude doesn’t. Claude Code knows your coding style, ChatGPT has never heard of it. Agents that book or buy things for you often know none of it.

This post explains where each assistant keeps your settings, why they don’t sync, and the three ways to fix it.

Why your settings don’t follow you

Every assistant keeps its own memory and its own instructions, in its own format. There’s no shared standard for “things about me,” so nothing moves between apps unless you move it.

Some apps let you import memory from another, but that’s a one-time copy. As soon as you change something in one app, they drift apart again.

Where your settings live, app by app

AssistantWhere it keeps things about youCan you edit it?Moves to other apps?
ChatGPT Saved memories, chat history and custom instructions Yes, in Settings › Personalization No; copy it by hand. It doesn’t import from other assistants.
Claude Memory, project instructions and profile preferences Yes, in Settings and in each project Can import from other assistants, as a one-time copy
Gemini Personal context: info you asked it to save, and past chats Yes, in Settings & help › Personal context No
Claude Code CLAUDE.md files and settings.json Yes, they’re files Only within Claude Code
Cursor and other coding agents Rules files and AGENTS.md Yes, they’re files Partly; some tools share AGENTS.md
These features change often. We check each row when we update this post.

Memory vs settings

Most AI memory is a note the assistant wrote about you. It’s useful for conversation, but it’s a guess, and you can’t always see how it’s used:

MemoryUser seems to prefer Italian food.

A setting is exact. An assistant can check it before acting, and you can see and change it:

SettingsFavorite cuisines: Italian, American, Mexican
Dining budget: $60 a person
Allergy: shellfish

For anything an AI does on your behalf, like booking, ordering or scheduling, settings are more reliable than memory.

What’s worth keeping portable

Three ways to make your settings follow you

  1. By hand. Keep a short “about me” block in a note and paste it into each app’s custom instructions. It’s free and private, but you update every app yourself, and agents usually can’t read it.
  2. A memory tool. Several tools sync memory across assistants. They’re good at recalling conversations, but most store free text rather than exact settings.
  3. A shared preference layer over MCP. MCP is the standard most assistants now use to connect to outside tools. A preference server stores your settings once, and every connected app reads the same values, with permissions you control.

The shared layer we built: OpenPrefs

We built OpenPrefs for option 3, and it’s free. You save a preference once, and Claude, ChatGPT, Claude Code and other MCP apps read it. New apps start with no access, you choose what each can see or change, and every read and change is logged with undo.

Tell one assistant, and the others already know.
Get started free See how OpenPrefs works

For builders: giving agents a user’s preferences

If you’re building an agent that books, orders or plans for people, it needs the user’s budget and rules before it acts. With OpenPrefs, your agent connects over MCP, or a REST API for agents that don’t speak MCP. It can search, get and set a user’s preferences and list items, within the permissions that user granted. Every call is logged for the user to see.

The endpoints, OAuth details and tool list are on openprefs.ai.

FAQ

Can I move my ChatGPT memory to Claude?

Yes, as a one-time copy. Claude’s memory import gives you a prompt to run in ChatGPT, and you paste what ChatGPT lists back into Claude. The two won’t stay in sync afterwards, and ChatGPT doesn’t import from other assistants.

Why does my AI keep forgetting my preferences?

Memory is kept per app, it can fill up or be rewritten, and the model decides when to apply it. Short, specific instructions are followed more reliably than long ones.

Is there a way to share one set of preferences across all my AI apps?

Yes, through a preference tool connected over MCP. OpenPrefs is one, and it’s free.

Is OpenPrefs free?

Yes.

Which apps work with OpenPrefs?

Claude, ChatGPT, Claude Code, Muse, Antigravity and any app that supports MCP. Agents that don’t speak MCP can use its REST API.

Is my data private?

New apps get no access until you allow it, and you choose what each can see or change. Preferences are encrypted at rest and in transit, and every read and change is logged.

How do I remove everything?

To cut off one assistant, disconnect it on the Agents page and its access ends at once. To remove everything, open Account on openprefs.ai and choose Delete account: it deletes every preference and list, your history and every connected agent, for good.