Claude and ChatGPT each remember some things now, but each keeps that memory inside its own app. The documents behind your answers live somewhere else again, so you still end up re-explaining yourself.
Underneath, large language models are stateless: built-in memory is each vendor's own layer on top, kept in that vendor's app. What is missing is a persistent memory layer you control — a place to store what matters to you and hand it back to whichever assistant you connect when it is relevant.
What is MCP?
The Model Context Protocol (MCP) is an open standard that lets AI assistants securely connect to external tools and data sources. Think of it as a universal plug: instead of every app inventing its own integration, an assistant that speaks MCP can talk to any MCP server.
One of the most valuable things you can connect an assistant to is a memory server — a store of your personal knowledge that the AI can read from (and write to) during a conversation.
How persistent AI memory works
With a memory layer connected over MCP, the flow looks like this:
- You store information once — preferences, documents, contacts, health details, or anything else worth remembering.
- The memory layer organises it automatically, so it is retrievable by meaning rather than exact keywords.
- Your assistant queries it when a conversation needs context — "What is my landlord's number?" or "What are my dietary restrictions?"
- The answer comes back grounded in your real data, not a guess.
The result: your AI finally knows you, and that knowledge persists across every new conversation, on every device.
Why not just paste everything into the chat?
You can, but it does not scale. Pasting context into every chat is slow, it burns tokens, and it disappears the moment the conversation ends. A dedicated memory layer is durable, searchable, and shared by every assistant you connect. JustOnce works with Claude, ChatGPT (Business, Enterprise and Edu), Cursor, VS Code, Gemini CLI, Codex and Claude Code. On ChatGPT Pro your assistant can read your memory but not save to it.
Where ChatGPT stands
JustOnce works in ChatGPT on Business, Enterprise and Edu, on the web: an admin adds it as a custom MCP server for the workspace. Save something in Claude and ChatGPT can recall it — one memory for both. Two limits are worth knowing: on a personal Pro plan OpenAI gives custom MCP servers read-only access (it can recall but not save), and through ChatGPT, JustOnce refuses to save what looks like a password, PIN, card, bank or ID number; add those in the app instead.
How justonce.ai fits in
justonce.ai is a persistent, intelligent memory layer built for exactly this. You store a memory once — a bill, a document, a preference — and our AI categorises it across 18 categories and makes it retrievable with natural language. Connect it to Claude with our connect guide, and your stored knowledge becomes memory the assistant can use in any conversation. The guide also covers ChatGPT, Cursor, VS Code, Gemini CLI and Codex.
Every memory body and uploaded document is encrypted at rest with AES-256-GCM before it is written — bodies under a key issued for their space, documents under a key for your account, with the keys held by justonce.ai — and hosted in Australia. JustOnce never uses your memories to train AI models, and our AI provider, Anthropic, doesn't train on them either.
Getting started
- Create a free account and add a few memories you refer to often.
- Connect justonce.ai to Claude using MCP.
- Ask your assistant a question that depends on your personal context — and watch it answer correctly.
Your AI does not have to start from zero every time. Give it a memory that lasts. See how it works or start free.
