Why ChatGPT and Claude don't share what they know
ChatGPT's memory lives in your OpenAI account, and Claude's lives in your Anthropic account. Each one only ever reads its own. So when you plan something in ChatGPT and then open Claude to build it, Claude starts from nothing: it doesn't know your goal, the decisions you made, or what's left to do.
That's fine for one-off questions. It gets painful when you use both for the same piece of work, which is common: many people like ChatGPT for thinking things through and Claude (or Claude Code) for writing and code.
The usual workarounds
- Copy and paste. Ask one AI for a summary and paste it into the other. It works, but summaries get long, details get lost, and you repeat it every time you switch.
- A notes document. Keep the plan in a doc and paste it in at the start of each chat. Better, but you're the one keeping it up to date.
- Project files. Upload the same file to a ChatGPT project and a Claude project. Each AI can read it, but neither can update it, so it goes stale.
All three have the same weakness: the AIs can read your notes but can't write back, so you do the syncing.
A shared memory both AIs can read and write
Both ChatGPT and Claude support MCP (the Model Context Protocol), the open standard for connecting AI apps to outside tools. A memory server connected over MCP gives each AI a few tools (save, search, load, add an update) that work on the same store. Whatever one AI saves, the other can load.
That's what Relay is: one private memory, connected to ChatGPT, Claude and any other AI you use. When you finish planning in ChatGPT, it saves a handoff brief (the goal, decisions, open questions and next steps). In Claude you say "pick up my latest plan from Relay", and it continues from exactly there. When Claude is done, it adds a progress update, and ChatGPT can review it.
Set it up in a few minutes
- Sign in to Relay with Google (free during the beta).
- In the dashboard, open Integrations and generate a link for ChatGPT, then follow the ChatGPT steps: in Settings, turn on Developer mode and create a connector with your link.
- Generate a second link for Claude and add it as a custom connector (Claude steps).
- In ChatGPT, plan as usual, then say:
Save this plan to Relay for the project "Web store".
Then, in Claude:
Pick up my latest plan from Relay and continue it.
What each AI can and can't do
- Each AI gets its own key, so you can see which AI wrote what and switch one off without touching the others.
- Facts you tell one AI to remember ("I prefer TypeScript") are available to all of them.
- AIs can't delete anything unless you allow it for that key, and even then only what they saved themselves, after asking you.
- Your memory is private to your account. Relay doesn't train AI models on it, show ads or sell data.