Agnos
A local document cleaning tool that prepares clean documents for AI. It supports PDF, Word, Excel, PowerPoint and web pages, removes noise such as headers, footers, page numbers and cookie banners, preserves tables, code and numbers, provides rule-level deletion transparency, and exports formats such as JSONL chunks.
Timeline (last 90 days)
Agnos is a product launched on Product Hunt, with the tagline "Clean documents for AI, and see what was removed," used to clean documents for AI and show what was removed.
Agnos supports cleaning documents in PDF, Word, Excel, PowerPoint, and web page formats.
Agnos mainly removes duplicate headers and footers, page numbers, cookie banners, boilerplate text, and duplicate content.
Agnos explicitly preserves structured information such as tables, code, and numbers.
Agnos labels each deletion with the name of the rule behind it, and users can keep a certain passage or simply turn off the corresponding rule.
Agnos runs on the user's Windows PC and does not upload any content.
Agnos supports export to JSONL chunks, plain text, or training data formats.
Agnos is distributed through the Microsoft Store, offers a 14-day free trial, and thereafter is a one-time buyout price of $99, with no subscription model.
Agnos has just landed on Product Hunt, and the coverage of its rule library and its effectiveness on documents with complex layouts still need more real-world use to verify.
All Facts (6)
Agnos supports cleaning documents in PDF, Word, Excel, PowerPoint, and web page formats.
50%UnverifiedAgnos mainly removes duplicate headers and footers, page numbers, cookie banners, boilerplate text, and duplicate content.
50%UnverifiedAgnos explicitly preserves structured information such as tables, code, and numbers.
50%UnverifiedAgnos labels each deletion with the name of the rule behind it, and users can keep a certain passage or simply turn off the corresponding rule.
50%UnverifiedAgnos runs on the user's Windows PC and does not upload any content.
50%UnverifiedAgnos supports export to JSONL chunks, plain text, or training data formats.
50%