chenxiachan/textfold is a Python tool that folds papers, web pages, files, and AI agent responses into five levels of detail, starting from the source text. The latest release identifiable from the provided repository information is v0.10.13. What changed in that version is not recorded in the given information. The five levels are arranged so you can move back and forth between the source text and the outline when working through long material.
The point is not just to reduce a document step by step, but to check the source text and the summary together. It reads PDF, Word, PowerPoint, Excel, EPUB, images, Markdown, and more; you can keep the original PDF side by side to leave highlights and notes, or ask questions about a selected part or the whole document. Responses from Claude Code and Pi can be folded as well, and `python3 -m textfold get` lets you feed documents to an agent or a script level by level. Document collection, source-citation-based summarization, plugin connections, and local storage are all gathered into a single tool.
That said, no performance comparison figures can be confirmed in the README. So from this information alone, it is not possible to judge whether its document processing speed or summary quality is better than that of other tools. Reading by moving between the source text and the summary across five levels is certainly clear, but a performance advantage has not been proven.