The questions people ask about memory in Claude, ChatGPT and Claude Code, answered plainly. Then what changes when one memory works across your tools and your people.
Yes, when memory is on. Claude saves what it learns as you chat, and on paid plans it can also search your earlier chats when you ask. Each is a switch in Settings: “Memory from chats” and “Search and reference chats”. On Team and Enterprise plans the owner decides whether memory is available, and each member turns it on. Every Claude project keeps a memory of its own.
ChatGPT’s memory is in Settings, under Personalization. Depending on your account it is the newer memory, a summary ChatGPT keeps updating from your chats, or the older saved memories, a list of individual details, and you can switch between the two there. OpenAI sets no separate storage limit on what ChatGPT draws from your past chats. If it says memory is full, open your saved memories there and delete the ones you no longer need. Turning Memory off does not delete your past chats, and deleting a chat does not delete a memory saved from it.
It belongs to one assistant and one account. What Claude remembers, ChatGPT never sees, and the other way round. It is not shared with other people: a colleague’s assistant does not know what yours learned. And deleting a chat does not remove what was remembered from it, in either app.
Claude Code remembers through files it loads when a session starts. Your CLAUDE.md files are read from the folder you start in and every folder above it, so a session started somewhere else does not see them. Auto memory, the notes Claude writes itself, is kept per git repository and on that machine only, so a new machine or a cloud session starts without it. Only the first 200 lines (or 25KB) of its MEMORY.md load at the start. And auto memory can be off: the toggle is in /memory.
Vega is a connector (MCP) for claude.ai, ChatGPT, Claude Code and Codex: one memory, sorted by topic, that each of them reads. In claude.ai and ChatGPT the assistant saves work when it calls Vega’s save tool and the app allows it, and saved work can be recalled in a later chat, in whichever tool you open. Share a topic and your team’s assistants read the same context: one place your team adds to as it works.
One install command adds a hook to Claude Code and Codex that saves your sessions as you work, with nothing to remember to do. Codex asks you to trust the hook, and again after it updates. In Claude Code, a new session can start with that project’s recent work. The hook reads only that tool’s own conversation files, and one command removes it. Cursor connects too, and saves when you ask it to.
Mem0, Zep and Letta are for teams building their own AI agents and apps: they give those agents memory and context. Vega is for people using the assistants they already have: you connect it, and it keeps your work across those tools and the people you share it with.