What actually needs to move
People picture moving a chat as moving the scrollback: every message, in order, into a new window. That is the one part that does not matter. An AI continuing your job needs five things, and none of them is the transcript:
The five things that carry a job
- The goal, in the words it was set, not a paraphrase of it.
- What is done, the finished steps and their outputs.
- What is next, the exact step the job stopped at.
- Every decision, including the dead ends, so they are not retried.
- The receipts, the checkable record under the steps that ran through tools.
Carry those five and the receiving AI is not starting over or interviewing you. It is continuing.
The three ways to move a chat, ranked
1. Hand the job over (the record moves)
Say wrap up in the AI that holds the job. It hands over the five things as one record with a continue link. Paste the link and the pickup prompt into any other AI, and the job continues, chained to the same record. This is the only method where nothing depends on your memory of the job.
2. Write the brief yourself (the summary moves)
Goal, done, next, decisions, typed by hand into the pickup prompt. Works in every AI with zero setup, and it is the right emergency move. Its ceiling is your recall: whatever you forget to write down is lost, and you will not know what it was.
3. Paste the transcript (the scrollback moves)
The last resort. Transcripts are long, noisy, and full of abandoned branches, and the receiving AI still has to guess which parts were decisions and which were drafts. If it is all you have, paste it and say what the goal is in the first line. Then set up method one so it is the last transcript you ever paste.
Method one, step by step
- Wrap up where the job lives. The line is in the card below. The AI returns a handover: the record plus a continue link.
- Open the receiving AI. Claude, ChatGPT, Gemini, Grok, or an editor with the MCP configured.
- Paste the pickup prompt and the link. The receiving AI reads the record and starts at what is next.
- Let the chain grow. Every hop lands on the same job record, so a job that crosses three AIs still reads as one job.
The one line to save
"When I say WRAP UP, hand this job over with Zambo: the goal, what is done, what is next, every decision, with the receipts attached, and give me the continue link."
Say it before the cap, and the AI hands the job over as a record: goal, done, next, decisions, receipts, and a continue link. Paste this pickup prompt into the AI that takes over:
"Continue this job from its record: [paste the continue link]. The goal, what is done, what is next, and every decision are in the handover, with the receipts attached. Start from what is next, and keep the record going as you work."
Nothing set up where the job lives? Skip the connector for today: the paste box at zambo.dev/continue takes the wrap-up from any AI and hands back the same continue link, no account and no install.
Setup by AI, honestly
The whole setup, the wrap-up line with copy buttons for each AI, and the paste box live on one page: Continue in another AI.
What each AI can do today, stated plainly:
Claude: one connector slot, on every plan
Claude takes custom connectors on every plan, and the free plan includes one connector slot. Add Zambo as a custom connector (Settings, then Connectors, then add a custom connector pointing at https://zambo.dev/mcp), then paste the wrap-up line into your custom instructions so it is loaded in every chat. From then on, the one-word flow works inside Claude itself, and Claude becomes the easiest place to create a handover.
ChatGPT: pickup for everyone, creating on team plans
Continuing a job inside ChatGPT works for everyone, through the pickup prompt: paste the prompt, add the continue link, and ChatGPT continues from the record. Creating the handover from inside ChatGPT needs MCP write access, which lives on the Business, Enterprise, and Edu plans. On a consumer plan, create the handover in an AI that has Zambo connected, Claude is the easy one, and bring the link to ChatGPT.
Gemini: the paste path, with Saved info
Gemini takes the paste path. Save the wrap-up line in Saved info so the habit is written down where you will find it, and when a job matters, wrap it up in the AI that holds the job, then paste the continue link into Gemini with the pickup prompt. Gemini continues from the recorded job instead of from your retelling of it.
Grok: a custom connector, or the paste path
Grok takes custom MCP connectors: add a custom connector at grok.com/connectors with https://zambo.dev/mcp, and the job can be wrapped up and continued inside Grok itself. Keep the wrap-up line in your custom instructions either way. With no connector set up, the paste path still works: paste the continue link plus the pickup prompt into Grok, and it picks the job up from what is next.
Cursor, Claude Code, Codex, VS Code: config once
In the coding tools, add the Zambo MCP to the MCP config once. Every project after that can wrap up and continue, and the handover runs between the editor and the chat AIs in both directions, which is the route to take when a chat AI planned the work and the editor has to build it, or the other way around.
A real job that already made this move
Claude started it, ChatGPT finished it, one chain
On October 18, 2026, a job ran in Claude, then continued in ChatGPT from the handover instead of from a summary, and was accepted on the record the same day. Open both hops yourself: the first handover, and the continuation. The second AI started at what is next, and nothing had to be re-explained, re-decided, or re-done.
The honest limits of moving a chat
What it costs to stop waiting
Don't buy a second AI for the three bad days a month. Finish in the ones you already have.
Continuing a job is free on a normal day: the handover, the continue link, and the receipt checks cost nothing, and the public handover page is free to send and to open. On a heavy day, when the job is big and the record has to hold up, a $0.99 Day Pass covers 24 hours of the full machine, and the Zambo Pass at $29 a month is there for people who hand work over every week. Start where the job is stuck: hand it over on zambo.dev.
Questions, answered
How do I move a chat from one AI to another?
You do not move the chat, you move the job inside it. In the AI that holds the job, say wrap up: When I say WRAP UP, hand this job over with Zambo: the goal, what is done, what is next, every decision, with the receipts attached, and give me the continue link. Take the continue link it returns, paste it into the other AI with the pickup prompt, and the job continues from what is next, with its goal, decisions, and receipts attached. With no setup at all, write the brief by hand: goal, what is done, what is next, and the decisions, dead ends included.
Can I transfer my chat history between AI apps?
The history itself, no. Claude, ChatGPT, Gemini, and Grok do not open each other's chats, and an export is an archive for you to read, not a state another AI can resume. What transfers is the working record the history produced, and that turns out to be the part worth having: the goal, the progress, the decisions, and the receipts, packaged as a handover the receiving AI continues from.
What gets lost when I move a chat?
With a transcript paste, plenty: decisions blur into drafts, dead ends look like live options, and the stopping point is wherever the scroll ends. With a hand-written brief, whatever you forgot to write. With a handover, the record is built to lose nothing on the list that matters: goal, done, next, decisions, receipts. What never moves is the first AI's private working style, and no method moves that.
Does the receiving AI need the same tools as the first one?
No. It continues from the record with whatever tools it has, and its steps chain onto the same job. The receipts from the first AI stay checkable by anyone, including you, regardless of which AI continues. If a later step needs a tool only the first AI had, the record says so plainly, and that step waits for the right tool instead of being faked by the wrong one.
Is moving a chat the same as exporting my data?
No. An export is a file of everything, made for archiving and reading later. A handover is a continuation point: a compact record of one job, built so another AI can pick the job up and keep working from it, with the receipts attached. Export when you are leaving a service. Hand over when the work needs to keep going somewhere else, including back again.