October 2026 — first public release

Tools

The catalogue launched with working tools across five categories: PDF, Images, Documents, Developer and Utilities. Every published tool has an implemented engine behind it — there are no placeholder pages, because a thin page that ranks is worse than no page.

  • PDF — merge, split, rotate, delete pages, extract pages, reorder, compress, watermark, edit metadata, remove metadata, PDF to JPG, images to PDF, extract text.
  • Images — compress, resize, convert, crop, rotate, flip, strip metadata, the popular format pairings, and two target-file-size tools.
  • Documents — extract text and data from DOCX, XLSX, PPTX and ODT, and build a DOCX from text.
  • Developer — format, validate, minify and convert JSON, CSV, YAML and XML, plus Base64, URL, JWT, UUID, regex, hash, colour, timestamp and case tools.
  • Utilities — word and character counters, text cleaning, line sorting and de-duplication, password generation, QR codes, and conversion helpers.

Platform

  • Local-first processing: no tool uploads files, by default and by construction.
  • Statically prerendered tool pages with per-page metadata, canonicals and structured data.
  • Command palette search on ⌘K and Ctrl+K, matching intent rather than just tool names.
  • Light and dark themes, both designed rather than inverted.
  • File validation on extension, MIME type and real file signature before any parsing.

Not shipped yet

Listed here rather than implied, so nothing here is mistaken for a feature. The honest reason is given in each case.

  • Password-protecting a PDF. The PDF library available in the browser does not write encryption. Doing it properly needs a native component, and a weak implementation would be worse than none.
  • OCR for scanned documents. Furtu reads an existing text layer. Recognising text in a page image is a different and much heavier problem, and a poor version is actively harmful because it produces confidently wrong text.
  • Downsampling the images inside a PDF. The compress tool rebuilds document structure but does not resample embedded images, because that trades away resolution silently.
  • Workflows as a real editor. The architecture anticipates chained steps, but no runnable workflow editor has shipped.
  • Accounts, tiers and billing. The pricing page describes intentions, not products. There is no account system at all today.
  • Additional languages. The string layer is centralised for future translation, but only English ships.

How this list is kept

Entries are written when something ships, not planned. If a feature is in progress it belongs in the engineering report, and if it is an idea it belongs in an issue. A changelog that describes intentions is a changelog nobody can trust.