How to import your Claude chat history into Whizi

The short answer

Whizi imports a Claude export from Settings, under Account, where the item is called Import Chats. Pick conversations.json out of your Claude export, or pick the export .zip and Whizi decompresses only that one entry from it. Your browser parses the file and posts the conversations, which land as ordinary chats in your list. Claude is one of exactly two accepted import sources, alongside ChatGPT.

What arrives is the text of the conversation. For a Claude export, thinking blocks, tool calls and attachments are dropped, and only text blocks are imported. That is a deliberate rule, not a failure, so an imported chat can legitimately be shorter than the original looked.

Question	Answer
Which file do I pick	conversations.json, or the export .zip it sits in
Does my export get uploaded	No. It is parsed in the browser, and only normalized conversations are posted
What carries over	User and assistant text, conversation titles, original timestamps
What does not	Thinking blocks, tool calls, attachments
How big can it be	A raw .json is accepted up to 300 MB. The .zip path has no ceiling of its own
What if I run it twice	The same chats land once. Re-running is the intended recovery

What a Claude export turns into

Each conversation in the export becomes one chat. The rules Whizi applies while it converts them are worth knowing, because every one of them can make an imported chat differ from what you remember.

Rule	What you see
Only text blocks are imported	Thinking blocks, tool calls and attachments from the Claude export do not appear
Titles are trimmed to 120 characters	A long title arrives cut, not rejected
A conversation with no usable title becomes Imported chat	Several such chats share that name
A message over the content cap is truncated	One oversized message does not cost you the rest of that conversation
A timestamp before 2015-01-01, or in the future, is discarded	That row falls back to the time you ran the import
A conversation with no readable text is counted as skipped	It is reported to you rather than silently dropped

The last row is the one people misread. A skipped count is information about your export file, not a sign that something went wrong in Whizi. It counts conversations that hold no readable text at all.

How a large export is handled

You do not have to split a big export yourself. The importer slices it for you, against caps that are the same for a Claude export and a ChatGPT one.

Cap	Value
Conversations per request	25
Messages per conversation per request	400, with the remainder sent as append slices onto the chat the first slice created
Request body accepted by the import endpoint	2,000,000 bytes
Request budget the website actually uses	1,200,000 bytes, deliberately under the cap to leave room for the request envelope
Import requests per minute	20

The rate limit is the part that looks like a bug and is not. An import that hits it is not surfaced as an error: the client waits 20 seconds and retries, up to 4 attempts per batch. So a big Claude export that seems to stall for twenty seconds at a time, then moves again, is behaving exactly as designed. Leave it running. The rest of the rate limiting picture is in rate limits.

Running the import twice is safe

Every imported conversation is keyed by account plus source plus source id. Re-importing the same export, resuming after a dropped connection, and clicking Import twice all land the same chats once.

Two consequences follow from that key. The first is that the source is part of it, so a Claude import and a ChatGPT import on the same account are tracked separately and cannot collide with each other. The second is that when an export gives no stable conversation id, the key falls back to a hash of the transcript itself, so those conversations are still matched on a re-run rather than duplicated.

There is one behavior to watch for. Deleting an imported chat and then importing the same export again brings that chat back rather than reporting a duplicate. If you deleted an imported conversation on purpose, do not re-run that export unless you want it returned.

Your export file does not leave your browser

The export file itself is never uploaded. Your browser reads it, parses it, and posts only the normalized conversations. That is also why every format error appears instantly, before anything has been sent.

Once the conversations are in, they are treated like any other Whizi chat. Conversations and messages are kept until you delete them or delete the account. And on training: "Whizi does not use your prompts, files, conversations, voice transcripts, or generated content to train Whizi-owned AI models. Whizi does not sell that content as training data." The full picture is in what Whizi does with your data.

If the import is rejected

Almost every failed Claude import is the wrong file rather than a broken import. Two messages cover the common cases:

That file is not a ChatGPT or Claude export. Look for conversations.json inside the export zip. means the parser could not identify either export format in what you picked.

There is no conversations.json inside that zip. Unzip it and pick the file directly. means the zip you picked has no matching entry.

Every other import message, the condition behind it, and the fix are on chat import errors, which is the page to open if the wording you saw is not one of these two.