Files & media

Save a file and the text inside it becomes searchable — scanned pages and screenshots included (OCR). The file itself stays in your library next to its extracted text, and that text is Markdown your own AI can read over MCP.

Supported types

Source	What you get	Chrome extension	Share a link (iOS & Web)
PDF (link or file)	Every page searchable, scans included (OCR) Share or upload from iOS, drag-and-drop or paste on web	N/A	● Optimized
Images & screenshots	Text in the image becomes searchable (OCR), plus a visual description iOS share/camera, web paste or drop	N/A	● Optimized
Word & PowerPoint	Document text, searchable iOS share sheet or in-app import (.docx, .pptx); not yet from the web or Mac app	N/A	● Optimized
Your own notes	Your own writing, stored as Markdown; never rewritten by Sensefold Composer on every platform: a rich editor on web and Mac, Markdown with a formatting bar on iPhone, plain text in the extension	● Optimized	● Optimized

● Optimized: tuned for the site · ◐ Works: general capture · ○ Not yet · N/A: not applicable on this path · How capture works

How do I save a PDF or a document?

Files come in through the apps. The Chrome extension captures web pages, not files, which is why its column above reads N/A.

Web app and macOS app — drag PDFs or images into the window, or paste a file or an image anywhere. Word and PowerPoint files cannot be uploaded from the web or Mac yet. See Web & macOS.
iPhone & iPad — share from Files, Mail, Photos, or any app that can share a document, including Word (.docx) and PowerPoint (.pptx) files. Inside the app you can also snap a photo, pick images from your camera roll, or import a document. See iPhone & iPad.
A link to a PDF — paste the URL into the web app or share it from iOS. Sensefold visits the link for you, as it does any shared link.

What a save gives you

PDFs — the text of every page, scanned pages included. Long documents are split into page-aware sections, so search can point at the right page. When your connected AI searches over MCP, a match inside a PDF carries the page numbers of the section it came from — see what a search result carries.
Images and screenshots — the text in the image (OCR) plus a visual description. A whiteboard, a slide, or a screenshot of a receipt is searchable by what it says.
Word and PowerPoint — the document text, searchable.
Every file — a summary and tags generated on save (very short documents skip the summary), and the original file kept with the item.

Search runs over the full extracted text, keyword and semantic together, so a phrase from page 40 finds the document, not only its title.

What the Markdown looks like

In the whole-library export, each saved file is one Markdown file: YAML front matter, then the extracted text as the body. Choose saved original files when you export and the PDFs, Office documents, and images you uploaded are included under originals/, referenced from the item:

---
type: Document
title: ...
description: ...        # the AI summary, if you included summaries
tags: [...]
sources:
  - path: originals/... # the saved original, with filename, mime_type, size_bytes, sha256
---

A single item can also be copied as Markdown from the iOS app’s item menu.

Your own notes

Notes sit in the same table because they live in the same library. Write them in the web or macOS editor (rich text that stays Markdown underneath), in the iOS app, or from the extension’s side panel; shared text on iOS also lands as a note. Notes are indexed and can get AI tags, but they are never rewritten or summarized automatically by Sensefold — only agents you authorize can edit them, versioned and revertible. Writing and editing notes never spends credits.

Credits and storage

Saving a file is free. OCR, the summary, and tags spend credits; PDFs are priced per page.
At zero credits a file is still saved with its original content; OCR and the summary wait. Once you have credits again, Process now in the web app processes pending items in bulk. For PDFs the cost preview is a lower bound until the page count is known.
Storage: 2 GB on Starter, 20 GB on Pro, and 200 MB without a plan.

Size and page limits

Uploads: up to 10 MB per file, for every supported type.
PDF links: the PDF behind a link can be up to 25 MB.
PDF pages: up to 200 pages. A longer PDF is still saved, but it is not processed: no text, OCR, or summary, and no credits are spent. Split it into parts of 200 pages or fewer and save those instead.
Word and PowerPoint files have no page or slide limit beyond the 10 MB upload size.

What is not captured here

Files through the Chrome extension. Use the web app, the macOS app, or the iOS share sheet.
Attachments inside an AI chat. Capturing a ChatGPT or Claude conversation keeps the conversation text, not the files you uploaded to it — see AI conversations. Save the file itself here.
Files on GitHub. A GitHub file link is handled by the GitHub capture: text files only, up to 1 MB — see Developer sources.
Your screen. The macOS app does not capture what is on your screen or in other apps. Take the screenshot, then paste or drag it in.

Troubleshooting

No summary or extracted text yet. The file was saved but enrichment has not run: either you are at zero credits (the item is pending) or AI consent has not been granted under You → Privacy & data. See Why did enrichment not run?
A file you just saved does not show up in search. Items become searchable after enrichment finishes, usually within moments.
An upload is rejected. Files over 10 MB cannot be uploaded; compress or split the file first.
A PDF was saved but has no text. Check the page count: PDFs over 200 pages are kept but not processed. See size and page limits.
A PDF link behind a login. Sensefold visits links as a signed-out visitor, so it cannot reach a file that needs your account. Download the PDF and drag it in or share it from your phone instead.

After saving

Everything here is readable by your own AI over MCP — ask Claude or ChatGPT about a report you saved and it can search the text and cite the item by its library link. Start with Sensefold for agents or the setup guides.