Keep a copy.
Keep your thinking.

Export without an AI connection

1. Open Settings → Saved data.
2. Find a saved document and choose its Export saved data action, or choose Export all… for every saved document.
3. Choose JSON or Markdown.
4. Select Export… and choose where to save the file.

Export all includes saved documents outside the recent-file list, including those whose original PDF has moved or is missing. Export needs no parsing, indexing, or provider connection.

Pending note edits and reading positions are saved across open windows before collection. Export includes chat messages already saved at that point. Let an active answer finish first if you want to include it. Unsent composer drafts and changes made after collection are not included.

If saving fails, retry with more disk space or a writable destination. Canceling the save dialog writes nothing, and a failed write leaves an existing destination unchanged.

Delete one document’s saved data

In Settings → Saved data, choose a document’s delete action and confirm. Platypus closes its open tabs and removes its highlights, notes, reading progress, chats and prepared AI data. Unsent chat drafts for that document are discarded. Your original PDF, other documents and shared model resources stay. Reopening the PDF starts fresh saved data.

Choose the format for your next step

Markdown is a single UTF-8 .md file organized by document. Readers with Markdown, footnote, and LaTeX support can render more formatting; an ordinary text editor can still read it. Exact colors, fonts, note sizes, and page positioning are not visually reproduced.

JSON uses export format version 1. The full format reference documents record fields, page numbering, geometry, and citation references.

Know what the file contains

Both formats include saved highlights, notes, conversations, attached selections, and citation references. They exclude original PDFs, settings, credentials, downloaded models, and retrieval indexes.

Exports are unencrypted. Sharing an export shares the excerpts, notes, and conversations inside it. Store exports wherever you would safely keep the corresponding document.

Where Platypus keeps your work

Annotations are app data; they are not automatically embedded in your PDFs. Platypus uses an Electron application profile beneath these standard locations:

Look for the app’s named directory, Platypus or platypus, depending on the build. Development and packaged builds may use different profiles. Preserve the active profile used by your build.

The profile includes reading state and annotations in Chromium/IndexedDB storage, chats under document-intelligence, settings, parsed text, indexes, and downloaded local resources. Original PDFs remain separate files.

Make a manual local snapshot

There is no built-in portable backup/restore flow. The current manual procedure is:

1. Fully quit Platypus so its databases are closed.
2. Copy the entire active application profile directory to a private backup location. Copying only chats.json or settings.json omits annotations and other state.
3. Separately back up your original PDFs. A profile backup is not a PDF backup.

Keep the snapshot private: parsed text, selections, and chat history can contain sensitive document content. Keep readable exports as an additional way to access your work.

Understand recovery limits

To attempt recovery, first preserve the current profile. With Platypus closed, restore the snapshot to the same profile path using the same app version and OS. Restore original PDFs to their previous paths where possible.

This is a filesystem recovery procedure, not a validated cross-device migration. Chromium storage can depend on the app’s origin and profile. Changing versions, OS, or development/packaged builds may prevent reuse, and downloaded native resources are platform-specific.

Keep the original snapshot until you have reopened PDFs and verified your notes, highlights, and chats. Legacy renderer-owned chats removed by IndexedDB v6 are not migrated.