How to keep AI memory across ChatGPT, Claude and other tools

Each AI tool keeps its own memory, stored in its own account, and none of them reads the others. To carry context across tools, keep it somewhere you control: a notes file you paste in, an AGENTS.md or CLAUDE.md file in your project, a memory server, or a shared node that every agent can read.

Why memory does not work across AI tools

A memory feature is a product feature, not a standard. ChatGPT saves memories in your OpenAI account. Claude saves them in your Anthropic account. Each company decides what to save, how to show it, and whether you can take it out. There is no shared format, so a tool cannot read what another tool learned about you.

Even inside one tool, memory has walls. Anthropic says each Claude project has its own separate memory space, and chat search does not cross project boundaries. So the fix is almost never a setting. It is keeping the important context in a place that is not owned by one product.

Option 1: use each tool's own memory

This is the easiest start. In ChatGPT, open Settings, then Personalization, then Memory to see and manage what it saved. In Claude, memory is under Settings, and you can view and edit individual topics, pause it, or reset it. Claude also offers incognito chats that are not saved to memory.

The limit is that this only works inside that one tool. It also saves what the tool thinks matters, not what you need. Read the list now and then and fix it.

Option 2: a notes file you paste in

The simplest cross-tool method is a plain text file you own. Write down who you are, what you are working on, your preferences, and decisions you do not want to revisit. Paste it at the start of a new chat in any tool. It takes a minute, works everywhere, and you can read exactly what the assistant knows. The cost is that you have to remember to paste and to update it.

Option 3: AGENTS.md and CLAUDE.md files

If you work with coding agents, put the context in your project. AGENTS.md is described on its own site as a simple open format for guiding coding agents, and it is supported by tools including OpenAI Codex, Cursor, VS Code and GitHub Copilot. Claude Code reads CLAUDE.md files, and Anthropic's docs say it can also read a repository's AGENTS.md, on its own or next to CLAUDE.md.

Anthropic also notes that Claude treats these files as context, not as enforced rules, and that short, specific instructions are followed better. These files are for code projects. They do not help a normal chat in a web app.

Option 4: a memory server (MCP)

Some tools connect to an outside memory store through MCP, a standard way for an assistant to use other programs. You run or sign up for a memory server, connect each assistant to it, and they read and write the same notes. This is the first option where memory is truly shared, but it takes setup, and you need to trust whoever runs the server, because it holds your notes.

Option 5: a node

A NodePress node is one version of option 4. It runs on your own computer. Agents join with one pasted line, and in the node they share a thread, a memory graph, and the files and context you choose to share. Every step is written to one readable thread, so you can see what each agent saw and said.

A node does not copy what a tool already saved in your ChatGPT or Claude account. You put in the context you want shared, or the agents add it as they work.

Which one should you pick?

Use a single tool and want less typing: its own memory. Use two or three tools and mostly chat: a notes file. Write code with agents: AGENTS.md and CLAUDE.md. Run several agents together on one task, or want one record of what happened: a memory server or a node.

Do it with a NodePress node

  1. Create a node on your computer and name it for the project or for yourself.
  2. Write the context you want shared, such as your role, preferences and decisions, into the node's shared memory.
  3. Paste the one line into each agent you use so they all join the same node.
  4. Tell each agent to read the shared memory first and write new decisions back.

If you use one assistant, or only need to paste a few lines now and then, a notes file or the tool's own memory is simpler. A node is useful once you have several agents and keep re-explaining the same background.

Questions people ask

How do I manage memory across multiple AI tools?

Keep your important context in one place you control, such as a notes file, project instruction files, or a shared node, and give it to each tool at the start.

Why doesn't memory work across AI tools?

Each company stores memory in its own account in its own format. There is no shared standard, so one tool cannot read another's.

Is there AI memory between chats?

Yes in many tools. ChatGPT and Claude both have memory features, but they work inside that tool, and Claude keeps each project's memory separate.

Does Claude Code read AGENTS.md?

Anthropic's docs say Claude Code can read a repository's AGENTS.md, on its own or alongside CLAUDE.md.

Can I back up my AI memory?

Yes. Claude lets you ask it to write out its memories verbatim, and ChatGPT shows saved memories under Settings, Personalization, Memory. Copy them into a file you own.

Related

Sources

NodePress gives your AI agents one shared place on your own computer. Free to start, no sign-up. Make your first node.