A document converter that does the work in your browser, and stays out of the way.
A document arrives as a .docx, a deck, a spreadsheet or an export zip, and the thing you want out of it is the words. Not the fonts, not the page breaks, not the twelve years of tracked changes — the words, in a form you can paste into a pull request, hand to a static site generator, diff, search, or read on a machine that does not have the program that made it.
Getting there usually means uploading the file to a stranger, installing something, or pasting the text and losing the tables. TransformPipe is the fourth answer: the conversion happens in the page you already have open.
Twenty conversions. A Word file, a slide deck, a spreadsheet, an EPUB book, an OpenDocument or rich text file, an Evernote export, a page of saved HTML, a CSV, a JSON response, or the zip that Notion, Confluence and Obsidian hand you when you export — each of them becomes Markdown. And Markdown becomes a finished HTML page, a Word file, a Confluence page, a Jira ticket, a Slack message, an EPUB book, plain text, or something you can print.
Everything normalises to Markdown, because Markdown is a format you can read without the tool that made it, diff in a pull request, and still open in twenty years.
The pictures come with it. A conversion that finds an image inside the file carries it into the output rather than leaving a reference to a folder you do not have, so what you get is one file that still shows what it showed.
Somebody moving documentation out of a wiki and into a repository, where the next person to change it can do so in a pull request instead of a rich text editor.
Somebody who works with an AI assistant all day and wants what it writes kept somewhere real — as a page with an address, rather than scrollback in a conversation that ends.
Somebody handed a file in a format they do not want to install a program to read, who needs the contents now and not after a download.
And anybody whose documents are not theirs to upload: a contract, a patient note, an unannounced set of figures, anything covered by an agreement that names who may process it.
Signed out, no file is sent anywhere. There is no upload to trust because there is no upload: the conversion happens on your own machine, which is why a page behind your company login converts as easily as one that is public.
This is a claim you can check rather than believe. Open the network tab, convert a file, and watch that nothing carrying it leaves — then turn the network off and convert another one.
Sign in and the Markdown is kept in your account, so a document follows you to another machine. It stays private until you share it, and a shared link can be revoked.
The exported HTML has its styles, its pictures, its diagrams and its formulas inside it. It asks the network for nothing, which is what makes it open the same on a laptop with no connection in five years as it does today — and what makes it safe to email to somebody who will open it once and never think about it again.
Most converters are a server: you upload, it works, you download, and a copy of your document exists on a machine you do not control for as long as their policy says. That is the right trade for some documents and the wrong one for others, and the point is that it should be a decision rather than a default.
Most converters also read the text and stop. A .pptx is a zip of XML parts, and the speaker notes — usually the most useful prose in a deck — are in a different part from the slides. An EPUB keeps its reading order in a spine that has nothing to do with the filenames. An Evernote export addresses its pictures by a checksum of their own bytes. Reading the container rather than the text is most of what separates a conversion you can use from one you have to repair.
And what it does not do is worth saying too: it does not read PDFs, LaTeX or the pre-2007 Office formats, it will not convert five hundred files while you make tea, and a single saved document stops at four megabytes. For those, Pandoc and the hosted services are the better tool, and the comparison of ten converters on the blog says which one for which job.
The same conversions reach a terminal, a pull request, a browser tab and a conversation: a public API, a command-line client with no dependencies, a GitHub Action that publishes the Markdown a pull request changed, a browser extension that converts the page you are reading, and an MCP server so an assistant can convert and share on your behalf. One set of converters behind all of them, so a table comes out the same wherever you asked.
TransformPipe is built by Raudar Labs. The source is public and MIT licensed, which means the conversion you are trusting is one you can read.