如何把 Claude 長對話轉移到 ChatGPT，避免重新交代內容

You are 80 messages into a Claude chat. The plan is half built, the decisions are made, and then you hit the usage limit. Or you want ChatGPT's take on the same problem. Either way, you open a new chat and it knows nothing.

Most people handle this in one of three ways. None of them is great.

1. Copy and paste by hand. Fine for five messages. For eighty, you lose the thread, the formatting breaks, and you spend ten minutes on it.

2. Ask the old chat for a summary. Works if the chat still has room. If you hit the limit, you can't. And a summary leaves out the small decisions you made along the way, which are exactly what you'll have to re-explain later.

3. Export the whole account. ChatGPT and Claude both offer data exports, but they give you a giant archive of every chat. You can't paste that into a new conversation.

What actually works

What you want is the one chat, turned into a block of text that a different AI can read and carry on from. I'll call that a capsule: the messages in order, plus a short instruction telling the new AI to treat everything in it as already settled and to pick up from the last message.

I built a free Chrome and Edge extension, ContextPod, that does this in one click on Claude, ChatGPT and Gemini. Here is how it works.

Open the chat you want to move.
Click "Save this chat" in the small panel at the bottom right. The extension scrolls through the whole conversation. It has to, because Claude only keeps the messages near the screen in the page, and the first version of my extension saved just the last one or two for that reason.
Open the other AI and click Insert. Short capsules go straight into the box. Long ones are copied to your clipboard and you paste them yourself.
The new AI replies with one or two lines on where you are, and waits for your next message.

Everything stays in your browser. There is no account and nothing is sent to a server.

Tips whichever tool you use

Put the instruction at the top, before the pasted chat. "Treat this as settled, don't re-ask" saves a round of questions.
Trim the oldest messages if the chat is huge. Models follow the end of a long paste better than the start.
Keep the last reply from the old chat intact. It is the most useful part.

Try it

ContextPod is free: Chrome Web Store

I built it with Claude Code. I'd like to hear what breaks, especially on Gemini, which I've tested the least.