tomd is a tiny desktop app that turns PDFs, Word docs, spreadsheets, slides and images into clean Markdown — the most token-efficient format you can feed an LLM.
Attach a Word doc, PDF or slide deck to Claude or ChatGPT and a huge slice of the context window disappears into formatting overhead before the model reads a single idea.
Markdown is the format LLMs natively "speak" — minimal markup, structure preserved, dramatically fewer tokens. tomd converts your documents before they hit the chat box, so the model spends its context on your content, not your file format.
Same content, structure intact — headings, lists, tables and links preserved as clean Markdown. Conversion quality by Microsoft MarkItDown.
tomd doesn't reimplement anything. It's a thin desktop front-end that runs Microsoft's open-source MarkItDown CLI on your machine — the exact commands you'd type yourself, minus the typing.
Download the app and open it. On first launch, tomd checks your
device for the requirements — Python 3.10+ (or uv)
and the markitdown CLI.
Already have markitdown? tomd uses yours and never
asks. Missing it? One click installs
markitdown[all] into a private environment — your
global Python is never touched. You watch every command run live.
Drag files or whole folders in. Each file runs through the queue,
and the .md lands right next to the original.
Drop single files or entire folders — tomd recursively picks up every supported file inside and lines them up.
Flip one switch and dropped files convert instantly — one at a time through a sequential queue, with a Stop button when you change your mind.
Per-file status, a spinner on the active conversion, an overall progress bar, and toast notifications when something finishes — or fails.
Copy the Markdown straight to your clipboard, drag the
.md file out into any app, or reveal it in Finder /
Explorer.
No uploads, no accounts, no telemetry. Conversion runs as a local process on your machine. Your documents never leave it.
Transparent by design: the footer shows exactly which
markitdown binary is in use, and setup streams every
shell command it runs.
Everything Microsoft's MarkItDown handles, through one drop zone.
tomd doesn't bundle a converter. It finds — or installs — Microsoft's MarkItDown on your machine and shells out to it, file by file. That keeps the app small, the conversion engine always upgradeable, and the whole pipeline inspectable.
Files and folders queue up in the tomd window.
markitdown file.pdf -o file.md
One subprocess at
a time, on your device.
Saved next to the original. Copy it, drag it out, or reveal it.
Drag & drop conversion for PDFs, Office documents, images, audio, HTML, archives and more — with auto-convert, a sequential queue, live progress and per-file actions.
Paste a YouTube link, get the transcript as Markdown. Instead of handing an LLM a URL it can't read, hand it the actual content.
A text field that takes whatever you paste — raw HTML, rich text, markup — and turns it into Markdown on the spot.
Pick an output folder, enable MarkItDown plugins, wire up LLM-powered image descriptions and OCR — the advanced knobs, kept out of the way until you want them.
tomd stands on the shoulders of MarkItDown, an open-source project by Microsoft's AutoGen team. The GUI, the issues, the roadmap — all public.
markitdown in a terminal. tomd has no server,
no accounts, and sends nothing anywhere.
markitdown isn't already on your PATH, tomd
creates a private virtual environment in your user library (e.g.
~/Library/Application Support/tomd/venv on macOS) and
installs markitdown[all] into it. Your system Python
and global packages are never modified. Every command is streamed
to the screen as it runs, and you can delete that one folder to
undo it completely.
markitdown binary is found, tomd uses it and never
installs anything. The footer of the app always shows which binary
is in use.
report.pdf becomes
report.md in the same folder. From the app you can
copy the Markdown, drag the file anywhere, or reveal it in Finder
/ Explorer.