TextFold

TextFold (formerly PaperFold) is an open-source "folding" reader: it folds papers, web pages, documents and even AI agents' replies into five levels of detail, so you can zoom between the full text and a few-word map of topics.

It is published on GitHub by chenxiachan (chenxiachan/textfold, Apache-2.0, Python 3.9+, runs on your own computer). The repo was created on October 4, 2026, has about 600 stars, and already has a third-party Obsidian bridge plugin. On October 25, 2026 it was featured in a Chinese "open source · product · AI" roundup on X, described as folding long text into five levels to avoid AI summaries that drop conditions and make it hard to find the original.

Five levels of detail

Level	Each paragraph becomes	In whose words
1 Full	the paragraph as written	the author's
2 Brief	its 1–3 key sentences, verbatim	the author's
3 Key	its point and the reason for it	AI, resting on the Brief sentences
4 Takeaway	one line	AI
5 Topic	a few words; the text becomes a map of section cards	AI

Because the first two levels keep the author's own sentences, you can always trace a summary back to the source.

What it reads

• Links: any web page, arXiv, PubMed Central, bioRxiv, medRxiv, Wikipedia, GitHub READMEs
• Files: PDF, Word, PowerPoint, Excel, EPUB, HTML, CSV, pictures, Markdown, text
• Agent output: a Claude Code mod and a Pi extension fold every reply one key per level, with no second model

Other features

• Words that stay slide to their new place as you zoom
• Read beside the original PDF pages
• Ask AI about a selected passage; answers cite the paragraphs they rest on
• Highlights and margin notes persist as you zoom out
• Sentence-by-sentence translation, ⌘K library search, Zotero and Obsidian support

How to use it

• Try the demo: gallery at chenxiachan.github.io/textfold
• Run locally: download the app from GitHub chenxiachan/textfold, then paste a link or drop files to open them as a map of sections
• In Claude Code / Pi: install the mod/extension and fold agent replies one key per level
• For agents and scripts:

python3 -m textfold get

hands a document to your agent or script one level at a time; an open format lets other programs read and write folded text.

Who it's for

• Researchers and students who read many papers and long documents
• Developers who want to skim the structure of long agent replies before diving in

Details: GitHub chenxiachan/textfold.