Open your ChatGPT export in the browser

Drop the zip OpenAI sent you. Read and search every conversation, download any one as Markdown or all of them at once, print one to PDF, or import them as pages. Nothing is uploaded: the file is read on your computer and never leaves it. Claude's export works too.

Drop your export here

The zip ChatGPT or Claude emailed you, or the conversation files inside it.

Read on your computer. Nothing is uploaded.

Getting the export, and what to do with it

01
Ask ChatGPT for your data
Open your profile menu, choose Settings, then Data controls, then Export. Confirm, and OpenAI emails a link to a zip. It can take up to seven days to arrive, and the link stops working 24 hours after it does, so download it the day it comes.
02
Or ask Claude for yours
In Claude, open Settings, then Privacy, and choose Export data. The link arrives by email in the same way. Both exports are account-wide: every conversation, not one.
03
Drop the zip here
Unzipped or not. Only the conversation files are read, so the pictures in the archive cost nothing: conversations.json, or the numbered conversations-0.json parts a large export comes in, which read as one list. The list fills with every conversation, newest first, and a search box finds the one you came for.
04
Take what you want out
One conversation as a Markdown file, all of them as a zip of Markdown files, one printed to PDF through your browser's own dialog, or all of them as pages in a vault you can search later.

What it reads

Every conversation in the export, with its title, dates and the model it used
The branch you last saw, the way the app shows it: an edited prompt or a regenerated answer leaves its old version in the file, and the viewer follows the one you kept
Code blocks, lists, tables and links, as the Markdown they were
A large export's numbered parts, conversations-0.json and up, as one list
Claude's export as well as ChatGPT's. The file says which it is

What it leaves out

Images and files. The export holds pointers to them, not the pictures, so the viewer says where one was and moves on.
Tool calls, browsing, code the model ran and its results. They are counted on a line under the title rather than shown.
Reasoning and thinking blocks, counted the same way.
Conversations you deleted before asking for the export. They are not in the file.

Questions

Is my export uploaded anywhere?: No. Your browser reads the file on your computer, and the conversations stay in this tab until you close it. The only button that sends anything is Import, which puts the conversations you chose into your own Sumibako vault and nowhere else. You can check this in the browser's network panel: nothing leaves while you read, search, download or print.
Does it read Claude's export too?: Yes. Claude's export is also a zip with a conversations.json in it, in a different shape, and the viewer recognises which one it was given. Grok, Gemini and other exports are not read yet.
My export has numbered files instead of conversations.json. Which one do I drop?: All of them, or the zip as it came. OpenAI splits a large export into conversations-0.json, conversations-1.json and so on, and the viewer reads them in order as one list. If you have unzipped it, select every numbered file at once.
Why is a conversation shorter than I remember?: Two reasons. ChatGPT keeps every edited prompt and every regenerated answer as a branch in the file, and the viewer shows the branch you last saw, as the app does. And tool calls, browsing, code the model ran and its reasoning are left out and counted in a line under the title, so a chat that was mostly the model working will read short.
Can I export one conversation straight from ChatGPT instead?: No. OpenAI's export is the whole account, arrives by email within up to seven days, and has no option for a single chat. Share links, printing and copy and paste are the alternatives, and each keeps something different; the post linked below goes through all of them.
What do the Markdown files look like?: A heading with the title, then the dates, the model and the chat's own address, then each turn under a bold You or ChatGPT label. They open in Obsidian, Notion, VS Code or any editor, and the zip holds one file per conversation, named after its title.
Does importing into Sumibako cost anything?: No. The free plan holds up to 1,000 notes, and each conversation becomes one page under a parent page, newest first. If the export is larger than the room you have, the import stops at the limit and says how many did not fit. The whole import can be undone in one step from Settings.

Every way to get a conversation out of ChatGPT, with what each one keeps, is in How to export a ChatGPT conversation.

Keep the ones worth keeping

A Markdown file in a downloads folder is where conversations go to be forgotten. A page in a vault is searchable from any device, can be shared with a link, and is where the agents you use can write too.

Start writing - free What the free plan holds

Getting the export, and what to do with it

Ask ChatGPT for your data

Open your profile menu, choose Settings, then Data controls, then Export. Confirm, and OpenAI emails a link to a zip. It can take up to seven days to arrive, and the link stops working 24 hours after it does, so download it the day it comes.

Or ask Claude for yours

In Claude, open Settings, then Privacy, and choose Export data. The link arrives by email in the same way. Both exports are account-wide: every conversation, not one.

Drop the zip here

Unzipped or not. Only the conversation files are read, so the pictures in the archive cost nothing: conversations.json, or the numbered conversations-0.json parts a large export comes in, which read as one list. The list fills with every conversation, newest first, and a search box finds the one you came for.

Take what you want out

One conversation as a Markdown file, all of them as a zip of Markdown files, one printed to PDF through your browser's own dialog, or all of them as pages in a vault you can search later.

What it reads

Every conversation in the export, with its title, dates and the model it used

The branch you last saw, the way the app shows it: an edited prompt or a regenerated answer leaves its old version in the file, and the viewer follows the one you kept

Code blocks, lists, tables and links, as the Markdown they were

A large export's numbered parts, conversations-0.json and up, as one list

Claude's export as well as ChatGPT's. The file says which it is

What it leaves out

Images and files. The export holds pointers to them, not the pictures, so the viewer says where one was and moves on.

Tool calls, browsing, code the model ran and its results. They are counted on a line under the title rather than shown.

Reasoning and thinking blocks, counted the same way.

Conversations you deleted before asking for the export. They are not in the file.

Questions

Is my export uploaded anywhere?

No. Your browser reads the file on your computer, and the conversations stay in this tab until you close it. The only button that sends anything is Import, which puts the conversations you chose into your own Sumibako vault and nowhere else. You can check this in the browser's network panel: nothing leaves while you read, search, download or print.

Does it read Claude's export too?

Yes. Claude's export is also a zip with a conversations.json in it, in a different shape, and the viewer recognises which one it was given. Grok, Gemini and other exports are not read yet.

My export has numbered files instead of conversations.json. Which one do I drop?

All of them, or the zip as it came. OpenAI splits a large export into conversations-0.json, conversations-1.json and so on, and the viewer reads them in order as one list. If you have unzipped it, select every numbered file at once.

Why is a conversation shorter than I remember?

Two reasons. ChatGPT keeps every edited prompt and every regenerated answer as a branch in the file, and the viewer shows the branch you last saw, as the app does. And tool calls, browsing, code the model ran and its reasoning are left out and counted in a line under the title, so a chat that was mostly the model working will read short.

Can I export one conversation straight from ChatGPT instead?

No. OpenAI's export is the whole account, arrives by email within up to seven days, and has no option for a single chat. Share links, printing and copy and paste are the alternatives, and each keeps something different; the post linked below goes through all of them.

What do the Markdown files look like?

A heading with the title, then the dates, the model and the chat's own address, then each turn under a bold You or ChatGPT label. They open in Obsidian, Notion, VS Code or any editor, and the zip holds one file per conversation, named after its title.

Does importing into Sumibako cost anything?

No. The free plan holds up to 1,000 notes, and each conversation becomes one page under a parent page, newest first. If the export is larger than the room you have, the import stops at the limit and says how many did not fit. The whole import can be undone in one step from Settings.

Every way to get a conversation out of ChatGPT, with what each one keeps, is in How to export a ChatGPT conversation.