---
title: "Memory"
description: "How the built-in Memory tool works: personal long-term context that an assistant stores and recalls."
canonical_url: "https://meingpt.com/en/docs/platform/memory"
language: en
---

# Memory

The built-in **Memory** tool gives an assistant a long-term memory: it remembers information about you across individual chats and recalls it later. That way you don't have to re-explain recurring context – your role, your working style, fixed preferences – in every chat.

Memory is **personal per user**. Your memories are tied to your account, are not visible to others, and are not automatically shared with your team.

## How Memory works

### Memory is enabled on the assistant

Memory is a **built-in tool** and only works when it is switched on in the assistant. You enable it in the [Editor](/en/docs/platform/assistants) under **Tools → Basic tools**. If the tool is off, the assistant neither stores nor reads any memories.

![Tools section of the assistant editor with the Memory tool enabled under Basic tools](/images/assistants/memory-enable-tool-en.png)

Admins can additionally disable Memory organization-wide under `Settings > Tool Management`, or leave it out of the approved list. When that's the case, the tool isn't available in any assistant, regardless of the editor setting. See [Tool Access Policy](/en/docs/admin-guide/tool-access-policy) for how the settings interact.

### Your memories apply across all your assistants

Memories are tied to **your user**, not to a single assistant. Every assistant of yours that has Memory enabled draws on the same personal memory store. So what one assistant stores about you can also be used by another assistant that has Memory enabled.

### The assistant decides when to use Memory

Memory is not loaded automatically into every chat. Instead, the **model decides for itself** when to use it:

- It **searches** your memories before answering a question that might depend on prior context.
- It **stores** a new memory when a stable preference or long-term context comes up.
- It **updates** an existing memory when you correct earlier information.

You can see these actions directly in the chat as their own tool steps.

![Chat where the user asks the assistant to remember something – the "memory added" tool step shows the stored memory](/images/assistants/memory-chat-toolcall-en.png)

### Search understands meaning, not just keywords

Memory search works **semantically** – it finds relevant memories by their meaning, not only on an exact word match. A question about "my preferred tools" will therefore also find a memory that names a specific tool.

## Tool methods

When Memory is enabled, the assistant has access to these methods:

| Method              | What it does                                     |
| ------------------- | ------------------------------------------------ |
| `memory_add`        | Store a new memory                               |
| `memory_search`     | Search memories semantically                     |
| `memory_get_all`    | List all stored memories                         |
| `memory_update`     | Update an existing memory                        |

The assistant can **add, search, list and update** memories – but **not delete** them. Deletion is entirely up to you in Settings (see below).

## Managing memories

You manage your stored memories centrally on the **Memory page in Settings**. There you can:

- **view** and **search** all stored memories,
- **delete** individual entries,
- select several entries and **delete them together**,
- remove the entire store via **Clear all**,
- **import** memories from another AI (see below).

![Memory page in Settings with search, import, delete-selected and the list of stored memories](/images/assistants/memory-settings-overview-en.png)

The content of a memory is **not edited** in Settings. To change a memory, delete it here or let the assistant update it in the chat.

## Import memories from another AI

If you have already worked with ChatGPT, Gemini, or Claude, you can bring the context they stored about you into meinGPT — without rebuilding it from scratch. On the **Memory page in Settings**, open **Import**.

The import has three steps:

1. **Copy the prompt:** meinGPT gives you a ready-made prompt. Copy it and paste it into ChatGPT, Gemini, or Claude.
2. **Paste the response:** bring the other AI's answer back and paste it in (up to 16,384 characters). meinGPT extracts the memories and sorts them into categories (Profile, Working style, Active work, Team & contacts, Rules).
3. **Review and import:** go through the suggestions, edit them if needed, and uncheck anything you don't want to keep. **Import** finishes the process.

Only the memories you selected are stored. Up to **50 memories** are imported per run, and each individual memory can be at most **500 characters** long. If the other AI's response contains very long, verbatim-quoted passages, the import can fail - shorten the response or ask the source AI for a more concise summary, then paste it again.

![Import dialog on the Memory page with the copyable prompt for ChatGPT, Gemini, or Claude](/images/assistants/memory-settings-import-en.png)

## Best practices

- Store **stable preferences** and **long-term context** – not fleeting one-off details.
- Keep memories **short and self-contained** so each stays usable on its own.
- Do **not store secrets** or highly sensitive data (passwords, keys).
- Have existing memories **updated** on corrections instead of creating duplicates.

## Where to go next

### [Editor → Tools](/en/docs/platform/assistants)

Where you enable Memory and other basic tools on an assistant

### [Create an assistant](/en/docs/platform/assistants)

Step by step to your own assistant