Read, convert and compare markdown without uploading it
Drop a .md file or paste text. Rendering, HTML export and the diff all happen in your browser — drafts, meeting notes and internal docs never leave your machine.
Drop a .md file here, or click to choose
.md, .markdown or plain text · rendered locally · no upload
…or paste markdown into the box below. Ctrl+V works too.
Drop a .docx or .pdf here, or click to choose
Parsed in your browser with mammoth / pdf.js · no upload · legacy .doc is not supported
Paste the old and new versions, or drop two files. Comparison happens locally.
Why run markdown locally?
A markdown file is rarely a blog post. It is a design doc, a set of meeting notes, an internal runbook, a private journal entry, or a draft nobody else should see yet. Most online markdown viewers and converters upload the text to a server to render it. This one parses and renders in your browser, so the file is read from your disk and never transmitted — you can confirm it yourself with the network tab open.
Converting DOCX and PDF files
The DOCX / PDF → Markdown tab reads a Word .docx or a PDF locally and gives you editable Markdown plus a live HTML preview. Headings, links, lists and tables survive a DOCX conversion; PDF conversion extracts selectable text page by page. When a chart or image is detected, the tool adds an honest image/caption placeholder rather than inventing chart data. Scanned or image-only PDFs have no text to extract, so the tool warns you rather than inventing content. Legacy binary .doc files are not supported — save them as .docx first. Every output can be downloaded or copied without ever leaving your device.
Exporting Markdown as DOCX or PDF
The View & convert tab now exports real .docx and .pdf files from your markdown — headings (h1–h6), bold, italics, inline code, links, ordered and unordered lists, fenced code blocks and GFM tables are all interpreted from the same parse. Both files are built locally in your browser: no text is uploaded, and no server is involved. The .html export remains a standalone document with its styles inlined.
Experimental local OCR
Experimental, best-effort: the Find charts / OCR button can run Tesseract locally from vendored files to read text from rendered PDF pages. It works only when your browser can render the page to a canvas; if it cannot, the tool reports a clear placeholder/warning rather than pretending recognition happened. OCR output is often imperfect — always compare it with the original. Nothing is ever sent to a recognition service.
What "standalone HTML" means
The export is a single self-contained .html file: the rendered content plus the styles needed to make it readable, with no external stylesheet, no script and no dependency on this site. You can email it, open it offline in five years, or commit it next to the source. That is the difference between an export and a screenshot of one.
How the comparison works
The diff is line-based and uses a longest-common-subsequence alignment — the same family of algorithm git uses for text. Each row is marked added, removed or unchanged, with the line number on each side, so you can jump to the place the change happened instead of reading the whole document again. Long runs of unchanged lines are collapsed to keep the actual change visible.
When a diff is the wrong tool
A line diff is excellent for prose and configuration, and poor for a document that was reformatted wholesale — reflowing every paragraph makes every line differ and the output becomes noise. In that case compare a section at a time, or render both versions and read them side by side.
Why I built this, and why your data never leaves your computer
Getting things done is easy now. The AI tools are everywhere, and vibe coding has made building things something almost anyone can do. But none of that helps when the thing you are working on belongs to someone else — your own records, your family's, a friend's, a colleague's, a client's.
That was my problem. Every time I needed to handle something sensitive, I ended up doing it by hand, on my own machine, and it cost me hours. Over time those hours turned into a collection of scripts. I am an AI/ML engineer, a data scientist and a Python developer — not a web developer, and not a UI developer. With the help of AI I can now take an idea from those scripts and turn it into a working JavaScript web app, one I can host so anyone else can use it without installing anything.
What that means in practice
Everything on this site runs inside your browser. Nothing you paste, upload or convert ever leaves your own machine. There is no upload step because there is no server to upload to — the code that does the work is already on your computer, downloaded once like a picture.
You have probably read the phrase "zero data retention". I want to be blunt about it: that phrase is mostly meaningless. It only promises that a copy is deleted afterwards. The file still travelled across the network, and you have no idea who was listening on the way, whether the AI provider trained on it, or whether their systems were breached a month later. Deleting your copy does not un-send it.
Because nothing is transmitted here, you can use this site on a network you do not trust. Public Wi-Fi at a café, an airport, a hotel. It genuinely does not matter: there is nothing crossing the wire to be intercepted.
Verify it yourself — fifteen seconds
- Press F12 to open Developer Tools, and click the Network tab.
- Use the tool normally — do the thing you came here to do.
- Watch the Network tab: no new requests appear. The only entries are the page's own files and, if advertising is live, an ad request — never your file.
- Or simply turn off your Wi-Fi and keep using the site. It still works. A tool that uploads cannot do that.
About the site itself
I want to share this with everyone. Yes, the interface is vibe coded — but the ideas and the features are mine, and every tool exists because I needed it myself and could not find a version I trusted.
The domain name was not cheap, so the site is supported by advertising and sponsorship. Those ads are chosen to be relevant rather than intrusive, because a page that earns money by annoying you is not worth building. If ads are not your thing, block them. Nothing breaks — the tools do not depend on the ad network being reachable.
What the ads do support is the next thing I am building: a coding agent that automatically masks personal and company-identifying information, so you can use whichever AI provider you like without worrying about what leaves your machine. People are buying seven-thousand-dollar machines to host a model locally just to keep their data in. You should not have to. Precise detection and masking does the same job for free — and when it ships, it will make some of these pages unnecessary. That is the point. Thank you for supporting it.