The short answer
A session is a window, not a vault. When it closes, whatever lived only inside it closes too. Work run through the Zambo endpoint at https://zambo.dev/mcp does not live inside it: every call is recorded as it happens, step, saved result, and a verifiable receipt, at links that outlive any session, any logout, any app. A fresh session, in ChatGPT again or in a different AI entirely, opens the record, verifies the steps it will build on, and continues the job. Zambo is the verifiable work layer for AI; surviving session death is the plainest thing that layer does.
Scope, stated once and applied everywhere on this page: a receipt cannot be created after the fact for work nobody recorded. an agent receipt is a record of one ai agent tool call that anyone can check: it keeps the exact saved bytes and their sha-256 digest, and it proves the result was not changed since it was saved, not that it was right.
The manual recovery, honestly
Log back in and reopen the thread. Do this first. In most cases the conversation is intact in history, and for a chat-shaped job, rereading it may be all the recovery you need.
Scroll and salvage. Pull out the outputs you can still use: copy the final drafts, re-download the files the session produced, screenshot anything rendered that you cannot copy. This recovers the visible products of the work. It does not recover the state: which steps ran in what order, what each tool call returned, what was tried and discarded. A new session reading the transcript has to reconstruct all of that, and reconstructions drift.
Accept the hard case. If the thread itself is gone, deleted, wrong account, a temporary chat that never saved, then the unrecorded work is gone with it. No tool can issue proof today for steps nobody recorded yesterday, and a page that tells you otherwise is selling you a story. What you can do from here is make sure the next session's work is recorded from its first call, which is the rest of this page.
The receipt-backed path, step by step
- The work runs recorded. With Zambo added to the AI, every call lands outside the session the moment it runs: the step, its saved result, and its receipt, each at its own link. You get 20 free calls per day, so the insurance starts free.
- The session dies. The record does not notice. Logout, expiry, closed tab, dead laptop: the record sits on the endpoint, readable through the Today view and the job's handover, independent of any app being open.
- A fresh session reloads the state. In a new session, same app or a different AI, open the handover link. The session reads the goal, the steps, and the next steps as recorded, instead of interviewing you about what the old session did.
- It verifies, then continues. Before building on inherited steps, the new session checks their receipts at the public verify endpoint, accepts the handover, and works on. The job's trail now spans both sessions, with the seam visible.
The same record answers the neighbouring pains: a session that did not die but simply forgot is the AI that lost the thread, and a deliberate move to a different AI is switching AIs mid-task. The continuity pattern underneath all three is AI agent session continuity.
Set it up before you need it
Recovery you arrange after the loss is salvage; recovery arranged before it is a non-event. The whole setup is two habits. First, add Zambo to the AI you work in, so calls record themselves as they run. Second, package a handover at natural breaks, end of a stage or end of a day, so a pickup link always exists even if you never need it. Neither habit asks you to change how you work with the AI. They change where the work lives while you do it: outside the window, in the vault.
Questions, answered
I got logged out of ChatGPT. Can I get my work back?
The conversation, usually yes: log back in and the thread is normally in your history. The work state is a different question. Tool results, intermediate steps, and generated files existed as parts of that session's flow; from the transcript you can reread them, but you cannot hand them to a fresh session as checkable state. If the job ran through the Zambo endpoint, its steps were recorded outside the session as they happened, each with a receipt, and those are fully recoverable: they never lived in the session that died.
Where does the work go when a ChatGPT session ends?
For unrecorded work, nowhere you can reach: the session's working context dissolves, and what remains is the visible transcript in the app that hosted it. For work run through Zambo, each call's step, saved result, and receipt sit on the endpoint at their own links, readable through the Today view and the handover record, from any session, in any connected AI, after any logout.
Can a new ChatGPT session pick up where the expired one stopped?
A new session in the same app can reread the old thread if you paste or reopen it, but it starts from the transcript, not from the state. A session with Zambo added can do the stronger thing: open the handover or the continuity record, read the steps as recorded, verify the receipts on the steps it will build on, and continue from the recorded state. The expired session becomes a closed chapter of the job instead of the end of it.
What should I do before a long AI session so an expiry never hurts again?
Add the endpoint before the job starts, not after it breaks. With Zambo added to the AI, every call is recorded as it runs: step, saved result, receipt. Package a handover at natural breaks in the work, end of a stage, end of a day, so a pickup link always exists. Then a logout, a closed tab, or a dead laptop costs you a login, not the job. Starting is free: 20 free calls per day.
What does it cost to keep work recoverable like this?
Recording itself runs on the normal call terms: 20 free calls per day, then a $0.99 Day Pass for 24 hours of higher limits, or the Zambo Pass at $29 for 30 days, one payment, $290 for 365 days, and $19 for 30 days for $ZAMBO holders. Verifying a receipt is free and needs no account. People pay by card through PayPal; agents pay in USDC on Base through x402.