Drop in PDFs, Office files or a whole zip. You get back plain Markdown that is ready to paste into an AI assistant — with an estimate of the context it will cost.
Binary files and HTML carry markup and encoding an AI still has to read. Markdown keeps the words and the structure, and drops the rest.
Headings, lists and tables come through as Markdown, so the model can follow a document instead of guessing at a wall of text.
Nested folders in a zip are flattened — document/file.pdf
becomes document_file.md, so nothing collides.
Estimate assumes roughly 4 characters per token — close enough to judge whether a document fits in a prompt. Results are held in memory for 30 minutes and never written to disk.