Claude and ChatGPT can't share memory natively: neither reads the other's memory, and Claude's import from ChatGPT is a one-time copy, not a sync. To share memory between Claude and ChatGPT, point both at one source you control: a profile doc, a shared Google Doc, an MCP memory server, a cross-app memory app or a shared workspace. For most people, a profile doc is enough. Docs checked 2026-11-06.
Which way should you pick?
Pick by how often your context changes and who needs to write to it.
- Leaving ChatGPT for Claude: Claude's memory import. One paste.
- Using both, context changes monthly or less: one profile doc pasted into both apps' instructions.
- Both should read the same living notes: a Google Doc both reach through Google Drive connectors.
- Memory that updates itself everywhere, and you'll pay: a cross-app product such as MemoryPlugin.
- Comfortable with MCP: a remote memory server such as Supermemory. Local servers don't reach ChatGPT.
- Your "memory" is shared project work (decisions, files, tasks) several agents read and write: a shared workspace. That's Tokenrip, and Tokenrip is ours.
Does Claude have saved memory like ChatGPT?
Yes. Claude's memory is on by default for Free, Pro, Max and Ultra plans on the web, Claude Desktop and Claude Mobile, and Claude "saves memory as a set of individual topics as you chat" (Claude help). ChatGPT's memory can draw on past chats, saved memories, custom instructions, saved files and connected apps, depending on your plan and region (OpenAI help). Neither reads the other, so what you tell Claude stays in Claude.
How do the 6 ways compare?
| Method | Works in Claude | Works in ChatGPT | Stays in sync | Effort | Cost |
|---|---|---|---|---|---|
| 1. Claude memory import | Yes, all plans incl. Free, web and Desktop | Source only: you copy out of it | No, one-time copy | One paste | Free |
| 2. Profile doc in instructions | Yes, Instructions apply to all chats | Yes, custom instructions: 1,500 characters on Free and Go, 5,000 on Plus and up | No, you edit two copies | Low | Free |
| 3. Shared Google Doc | Yes, Google Drive connector for all users | Yes, Google Drive app; availability varies by plan and region | Yes for reading; editing depends on the app | Low | Free |
| 4. MCP memory server | Yes, custom connector (1 on Free) or local server | Remote servers only; full MCP on Business, Enterprise and Edu, read/fetch on Pro | Yes, if both connect | High | Anthropic's and OpenMemory free (self-hosted); Supermemory Free with $5 credits/month, Pro $19/month |
| 5. Cross-app memory product | Yes, extension or MCP | Yes, extension, Custom GPT or plugin | Yes | Low | MemoryPlugin from $89/year; Unabyss from $13/month billed annually |
| 6. Shared workspace (Tokenrip) | Yes, MCP connector | Yes, MCP connector; OpenAI's plan rules for custom MCP apply | Yes, when an agent loads it | Medium | Free: $0, 2,500 actions/month; Pro $20 per member/month |
Sources: Claude memory import, Claude personalization, ChatGPT custom instructions, Claude Google Workspace connectors, ChatGPT connected apps, Claude custom connectors, ChatGPT custom MCP servers, Supermemory pricing, MemoryPlugin, Unabyss, Tokenrip pricing.
1. How does Claude memory import from ChatGPT work?
Claude's memory import copies what another assistant remembers into Claude's memory in one paste. It's available on Free, Pro, Max, Ultra, Team and Enterprise plans, on the web and Claude Desktop (Claude help).
- In Claude, open Settings > Your Claude and select Start import. Claude shows an export prompt.
- Paste that prompt into ChatGPT. ChatGPT lists what it remembers in one code block.
- Paste the reply back into Claude and click Add to memory.
It's free and quick, but it's a snapshot. Anthropic calls imports "experimental and still in active development", and says Claude may not retain imported personal details unrelated to work. Going the other way, ask Claude to "Write out your memories of me verbatim, exactly as they appear in your memory," then paste the result into ChatGPT yourself.
2. Can one profile doc act as memory for both?
Yes. A profile doc is a short text (who you are, how you want answers, what you're working on) that you paste into Claude's Instructions and ChatGPT's custom instructions. Claude applies its Instructions to all of your conversations (Claude help), and ChatGPT applies custom instructions immediately to all chats (OpenAI help).
It's free, works on every plan, and you see exactly what each assistant is told. The catch: two copies to keep in step by hand, and ChatGPT's limit of 1,500 characters on Free and Go, 5,000 on Plus and up. Our opinion: for rules like "never use bullet points with me", this beats both built-in memories, because nothing gets rewritten behind your back.
3. Can Claude and ChatGPT read the same Google Doc?
Yes, through each app's Google Drive connector. Claude's Google Workspace connectors are available for all users on Claude and Claude Desktop, and Docs added to chats and projects sync from Drive, so Claude reads the latest version (Claude help). In ChatGPT, connect Google Drive from Settings > Plugins; availability depends on your plan, region and workspace (OpenAI help).
Put your profile and project notes in one Doc, connect Drive in both apps, and tell each app in its instructions to read that Doc first. You get one copy, current for reading. Writing back is uneven: Claude can edit Google Docs live in a pane beside the chat, while what ChatGPT can change depends on the app. More on this route in how to use Claude and ChatGPT together.
4. Which MCP memory servers work with both Claude and ChatGPT?
Only remote ones reach ChatGPT without extra plumbing: ChatGPT connects to remote MCP servers, and one on your own machine needs OpenAI's Secure MCP Tunnel (OpenAI help). An MCP memory server is a small service with tools like "add memory" and "search memory" that any MCP client can call.
- Anthropic's reference memory server is "a basic implementation of persistent memory using a local knowledge graph" (GitHub). It's free and runs locally in MCP clients such as Claude Desktop and VS Code. Not ChatGPT, out of the box.
- OpenMemory by Mem0 is a local MCP memory server with a dashboard. It runs in Docker, needs your OpenAI key, and keeps memory local "with no cloud sync or external storage" (Mem0). Free, with the same ChatGPT limit.
- Supermemory MCP is remote, at
https://mcp.supermemory.ai/mcp, and signs in with OAuth, so you don't paste an API key (Supermemory docs). The Free plan is $0 with $5 of credits a month, and Pro is $19 a month (Supermemory pricing).
Claude's Free plan allows one custom connector (Claude help). OpenAI says "Full MCP is only available to Business and Enterprise/Edu users, currently," and Pro users can connect custom MCP servers with read/fetch permissions. Our guide to adding MCP to ChatGPT covers where OpenAI's pages disagree. Looking for mem0 alternatives for your own chats rather than an app you're building? Option 5 fits better, since Mem0's hosted platform is an API for developers.
5. Is there an app that syncs AI memory across Claude and ChatGPT?
Yes. Cross-app memory products keep one memory outside both apps and feed it into each. Two current ones are built for people, not developers:
- MemoryPlugin connects through a browser extension, a Custom GPT that it calls the "fastest path for ChatGPT, also works in the iOS and Android apps," and an MCP server for Claude Desktop, Claude Code and Cursor. It imports chat history from ChatGPT, Claude, Gemini, TypingMind, Grok and DeepSeek. Core is $89 a year after a 7-day free trial (MemoryPlugin).
- Unabyss adds itself to Claude from its connector directory and to ChatGPT through a Unabyss plugin, and also pulls context from tools like Slack, Gmail and Notion. Pro is $13 a month billed annually for up to 3 connected agents via MCP (Unabyss).
These come closest to "tell Claude once, ChatGPT knows". The cost is another subscription and a third party holding your context, so read the privacy terms before importing years of chats.
6. What about a shared workspace like Tokenrip?
Tokenrip is ours. A shared workspace keeps memory as files every connected agent loads before it starts: a pinned profile, rules and active projects. Claude, ChatGPT and Grok chat connect to our workspace through the Tokenrip MCP connector. Claude was the easiest to connect, and ChatGPT works for us on web and desktop. Our ai-memory skill builds the three pinned files from your exported memories, writes what you say straight in, and proposes anything it inferred for your approval.
The cons: it isn't automatic background memory. Agents read and write it when told to or when a skill says so, so you add one line to each app's instructions ("Before answering, load my Tokenrip workspace me"). Every change an agent makes counts as an action against your plan's allowance; reading is free, and Free includes 2,500 actions a month. OpenAI's custom-MCP plan rules apply in ChatGPT. If you only need a profile, option 2 is enough.
How do you set up memory in Claude Code?
Claude Code keeps its own memory in files, in each project and under ~/.claude/projects/ (Claude Code docs). CLAUDE.md holds instructions you write, and /init generates a starting one. Auto memory saves Claude's own notes; it's on by default in local sessions, and the /memory toggle turns it off.
We had auto memory on, then turned it off. It applied stale decisions to current work, nothing in it decays, and it bloats context. We keep CLAUDE.md to about 150 lines as a router to task files instead (details). Claude Code is a third memory silo, so whatever source you pick has to reach it too.
Keep one memory that Claude and ChatGPT both read and update
Every option above either copies your context once or leaves you editing it in two places. Tokenrip keeps one memory, your profile, rules and active projects, in a workspace that Claude, ChatGPT and your other agents load first and keep current as you work.
- Tell Claude the launch moved to March, and ChatGPT's next draft says March.
- Your profile, rules and projects sit in three short files you can read and edit in a browser.
- Anything an agent guessed about you waits as a proposal until you approve it.
- Every change names the agent that made it and why.
- Adding a new assistant means one setup line, not re-explaining yourself.
Connect me to Tokenrip. Fetch https://tokenrip.com/setup?ref=claude-chatgpt-shared-memory and follow it. Then read the skill at https://tokenrip.com/skills/88ca6e75-6767-4c3e-ac5c-3fdd8cc4118d and set up one memory (my profile, rules and active projects) that Claude and ChatGPT both read and keep current.
Paste that into Claude, ChatGPT or any agent you use. The Free plan is $0 with no credit card: unlimited workspaces and agents and 2,500 agent actions a month.
Nothing leaves the workspace until you approve it: an agent that wants to send, post or publish proposes it first.
Related reading: