Think on a canvas. Leave with one clean markdown file.
The same paste. Two outcomes.
Pasted into a whiteboard
Flat text. Structure gone. Exports as a picture.
Pasted into Mark
- Paste markdown
- Keep the structure
A styled, movable card. Exports as text — in one ordered .md file.
Messy in, plain markdown out.
1Paste.
Markdown from an AI chat lands as a styled, movable card — not flat text.
2Spread out.
Arrange cards, sketch arrows around them. Be as messy as the thinking requires.
3Order.
Click the pieces that matter, in the order you want them. Badges track the sequence.
4Export.
One clean .md file — optionally run an AI pass: “turn this into a PRD.”
Write with AI. Export with AI.
Prompt a card onto the canvas. At export, say what the file should become — the AI formats your pieces, in your order.
Write with it
A prompt becomes a markdown card, ready to move and edit.
Export through it
One instruction reshapes the export — saved beside the raw file, re-runnable anytime.
Models fromAnthropicOpenAIGoogle
The board is not the product. The file is.
- 123→one .md
Your arrangement is the document order.
The sequence you set on the canvas is the sequence of the file. One document, your order.
- text outscreenshot
Cards export as text. Always.
An enforced rule in the export pipeline, not a setting. Your prose never becomes a screenshot of itself.
- instruction→formatted .md
The AI pass is built in.
No copy-paste-into-ChatGPT dance. The instruction runs on exactly the pieces you selected, in exactly your order.
- ObsidianvimGitHub
Plain markdown. No lock-in.
No wikilinks, no proprietary IDs. The file reads the same in Obsidian, vim, or GitHub — with or without Mark.
Fair questions.
Why not just use Excalidraw or Obsidian Canvas?
They export pictures of your board. Mark exports the document. If a picture of your board is the deliverable, use them — Excalidraw's sketching is excellent. But Excalidraw exports PNG/SVG/JSON with no markdown-card primitive, and Obsidian Canvas exports an image while its .canvas format is JSON. Mark exists for the moment the board needs to become a document.
Why not paste everything into ChatGPT and have it reformat?
That's the manual dance Mark removes. You'd be copying fragments one by one, describing the order in prose, and hoping the model reassembles what you already assembled by hand. In Mark the arrangement itself drives the export, and the AI pass runs on exactly the pieces you selected, in exactly your order.
Another app, another account — why?
Signup exists so canvases sync. Your work leaves as files. Email and password, with an email confirmation step; no guest mode yet. The account stores your scratchpads; the artifacts are files on your disk.
Is it free?
Yes. No plans, no tiers, no billing built. Mark is new and the current goal is simple: people using the paste-to-card-to-export loop for real work.
Is my data trapped if Mark disappears?
No. The export is the portable format. The artifact is a plain .md file with no Mark-specific syntax — self-contained with inline images, or a zip of the file plus assets. There's no proprietary format to migrate away from.
Is it powerful enough?
Honestly: Mark is deliberately small. A canvas comfortably holds hundreds of elements; snapping, alignment guides and rotation are built in, and real-time collaboration is planned but not built. If you need a general-purpose team whiteboard or a heavy PKM system, this isn't that. It's a scratchpad with a clean exit.
Who's using it?
It's new. No user counts, no logos — and we won't invent any. The proof on this page is product behavior, not testimonials: markdown cards export as verbatim text by an enforced rule, every export is a saved, reproducible record, and the output is plain markdown you own. Try it on one real document and judge it on that.
Be messy here. Ship something clean.
Paste your next AI answer onto a canvas instead of a doc. Put it in order, and export a markdown file you own.
Free. No card. Plain .md exports, yours forever. Already have an account? Log in.