What changed in Mark, in plain terms — new capabilities, fixes, and honest notes on tradeoffs. Newest first.
Pasting text onto the canvas now creates a card whose width grows with the content up to a sensible cap, then whose height grows to fit — no more internal scrollbar on a freshly pasted card. The toolbar's Paste as card button behaves the same way; Add card and AI-generated cards stay on their existing sizes.
The dashboard now has a folder sidebar: file thinkpads by dragging a card onto a folder, or from the card's menu via Move to. Folders can hold one level of subfolders, and everything still lives under All thinkpads until you file it.
You might keep a Client work folder with a subfolder per project, and leave personal scribbles unfiled.
Drop an arrow's end on a card, shape or icon and it snaps to that element — drag the card anywhere afterward and the arrow follows, stopping just outside its edge. Diagrams stay correct as you rearrange the board instead of drifting into stale, disconnected lines.
You might sketch a service diagram, then reshuffle the boxes into a new layout without re-drawing a single arrow.
Click away from a markdown card you never typed into and it's gone, instead of leaving an empty bordered rectangle behind. Cards you actually wrote in are never touched.
Connected agents can now create a canvas, rename it, and put it away again — so you no longer have to make and name an empty board before your agent can draw on it. Nothing is ever deleted: archived canvases come back from the dashboard's "Show archived" toggle.
You might say "make me a board for the payments architecture" and get a link back to a canvas that already has the diagram on it.
Editing a markdown card no longer has to double its width. A new button in the edit chrome hides the live preview pane, and Mark remembers your choice for every card from then on.
You might hide the preview once and never see the doubled-width edit view again.
A fresh welcome canvas now greets you with a big handwritten title and a small tip about pressing T, so the Text tool is impossible to miss on your very first open.
Connected agents can now put free-form text blocks on your canvas and style them — ink color, font, size, and alignment — instead of being limited to markdown cards.
You might ask an agent to add a big handwritten title above the cards it just created.
A new Text tool puts words straight on the canvas — no card, no markdown, just type. Pick an ink color, a font (sans, serif, mono, or handwritten), a size, and an alignment for any block, and it grows to fit whatever you write.
You might press T, click above a sketch, and add a big handwritten title in purple.
Press and hold on empty canvas and a small "Drag to select" label appears at your fingertip the moment the gesture takes over. Before, half a second of holding looked exactly like nothing happening.
You might rest a thumb on the canvas, see the label, then drag a box around three cards to select them all.
Dragging from the buttons and hint cards that float over your canvas now moves the canvas, instead of doing nothing. Tapping them still does what it always did.
On a phone the top strip and the round buttons used to be dead spots — put a thumb there and drag, and nothing happened. Now the canvas follows your thumb from anywhere.
If something interrupts you mid-drag — a system gesture, a swipe in from the edge of a phone screen — whatever you were moving, resizing or rotating now stops where you left it instead of following your cursor afterwards, and a half-drawn selection box no longer stays on screen.
Select an image — or several, plus any notes you want to include — and ask a question about them. The answer arrives beside your selection as an ordinary Markdown card you can edit, move and export.
You might select a screenshot of a chart and ask what it shows, and get the numbers written out as a card you can keep.
Select one or more images — and any note you want to guide it — then ask Mark for a new image. It arrives beside your selection as an ordinary image you can move, resize and export. Mark always shows you what it's about to make before it makes it, and free accounts include a set number of image generations each month.
You might select two product photos and a note about your brand colours, and get a single illustration that combines them.
Press and hold on an empty part of the canvas, then drag — a selection box follows your finger and everything it covers gets selected. A normal one-finger drag still moves the canvas around, exactly as before.
You might hold, draw a box around a cluster of cards and shapes, then move them all somewhere else in one go.
On a phone, the export panel now starts as a slim bar along the bottom of the screen showing how many things you have picked, so the canvas stays in view while you tap elements into order. Tap the bar to open the full panel when you want to rearrange or export.
You might sequence eight cards on your phone without the panel ever hiding the one you are reaching for.
Pens, shapes, arrows and icons can now be blue, green, amber or purple, alongside the greys and red you already had. Colour reads the same way in light and dark, so a diagram you colour-code at night still makes sense in the morning.
You might draw one subsystem in blue and the thing it talks to in green, so the arrows between them are obvious at a glance.
Mark has always had a light theme — now there's a control for it, on the dashboard and on the canvas. Pick light, pick dark, or follow whatever your computer is set to.
You might leave it on System so Mark goes dark with the rest of your screen in the evening.
Your canvas now shows its name in the top bar, next to the way back. Click the name to rename it — the new name appears straight away, and it is already there when you go back to your canvases.
You might open an untitled canvas, sketch until the shape of it is obvious, then name it on the spot.
Buttons, text fields and menu rows now grow to a comfortable tap size on any touch screen, so you stop missing the one you meant and hitting the one next to it. Nothing changes when you're using a mouse.
You might rename a thinkpad from your phone without two failed jabs at the menu button first.
On a phone, the edit button on a selected card used to shrink as you zoomed out, and once the card was small enough a tap could grab its resize corner instead of opening the editor. It now sits just above the card at a constant size, so it opens the editor at any zoom.
Dropping, pasting, or picking multiple images now adds all of them — before, only the first one landed and the rest were silently dropped. They arrive laid out in rows next to each other, all selected, and a single undo removes the whole batch.
You might drag a folder of screenshots straight onto the canvas and arrange them from there.
Hover any code block on a card and a copy button appears, putting just the code on your clipboard — no fences, no language tag, no hand-selecting. Selecting a card also gives you a button that copies the card's entire markdown.
You might paste a snippet from a card straight into your editor, or lift a whole card back into the doc you're writing.
Connected agents can now set a card's surface and border colour as they create it, so a board can mark what's a risk or a heading instead of arriving as a wall of identical cards.
Filling a card with a dark colour could leave its text invisible. Cards now offer a solid and a tinted surface — use the border colour when you want a card to stand out.
After your first card or sketch lands on a canvas, a one-time hint points at Export — the step that turns the pile into a single markdown file.
New accounts now start with a Welcome to Mark thinkpad already on their dashboard — a styled walkthrough of paste, sketch, and export — and land directly in it on first sign-in.
The check-email page now has a resend button with a cooldown, so a lost or slow confirmation email no longer strands new signups.
The Connect an agent dialog now shows every setup step — the connect command and the skill install — as soon as you open it, instead of hiding them until you create a token. Each command has its own copy button, so you no longer have to select multi-line shell commands by hand.
Alongside the MCP connection there's now an installable Mark skill — one command in the Agent tokens dialog — that teaches your agent Mark's craft: read the canvas before writing, tidy card grids, arrows that line up, and the palette. "Diagram this repo" comes out looking intentional.
Install it from the Agent tokens dialog, right under the connect command.
The same quiet animated canvas from the log-in page now sits behind the whole home page — pen strokes, drifting doodles and typed-out markdown snippets, sketching away while you read.
While editing a card, the two halves are now labelled Markdown and Preview, the preview side is gently tinted, and an empty card says "Preview appears as you type" — so the live preview reads as a preview, not a mysterious second box.
While you edit a card, the old selection frame no longer lingers at the card's committed width — it looked like a second phantom box beside the editor. The editing surface also gets a solid background and a friendly placeholder, so an empty card is easy to read on any canvas.
The log-in and sign-up pages now sit on a quiet animated canvas — pen strokes sketch themselves, snippets of markdown type out on little ghost cards, and small doodles drift past on a faint dot grid. It keeps out of the way and stays still if you prefer reduced motion.
Mark is now an MCP server. Create an agent token from the dashboard, connect Claude Code, Cursor or any MCP-capable agent, and it can read your canvases and lay out cards, shapes, arrows and icons for you — ask it to diagram a repository and the board fills in.
Cards, images, shapes, arrows and pen strokes can now be rotated: drag the round handle above a selection, and hold Shift to snap to 15° steps. Groups rotate together, and rotated sketches export exactly as you see them.
Double-click a card and the editor now shows your raw markdown on the left and the styled result on the right, updating as you type — no more committing blind to see how it looks.
An empty canvas now shows a small card with the five core moves — paste markdown, double-click to edit, number-key tools, Export mode, ⌘K to switch thinkpads. It vanishes the moment you start working, and dismissing it is permanent.
The export panel now reopens the way you left it — your last-used format (inline, zip, or links) and the standalone-arrows toggle stick between visits.
The AI generate dialog now always opens on top of the export-order panel — opening export mode while the dialog was open could leave it hidden behind the panel.
Cards, shapes and selections now snap to the edges and centres of things around them as you drag or resize, with guide lines showing what lined up — lining up a tidy row or column no longer takes careful nudging.
Shapes, arrows, pen strokes and icons can now be dragged to a new spot with the select tool — before, they could only be selected. Grouped shapes move together, and dragging on empty canvas still rectangle-selects.
Select several shapes and press ⌘G (or the new Group button) to make them one unit — it moves together, and exports as a single combined image instead of separate fragments.
You might group a sketched flowchart so the export shows one clean diagram.
The toolbar is now numbered 1–9 in order, and each button shows its number in the corner: 1 select, 2 rectangle, 3 ellipse, 4 arrow, 5 pen, 6 eraser, 7 image, 8 generate with AI, 9 export. Press a number and the matching tool or action fires.
The prompt box in the AI dialogs is bigger, and it now scrolls instead of stretching the dialog off the screen when you paste a lot of text. Size limits on what you can send to the model are gone too — no more “too long, shorten it”.
The “Generate a card with AI” and “Edit card with AI” dialogs could cut off their text box and buttons at the right edge. Everything now fits — and the dialogs are a little roomier.
Select a Markdown card and click its AI button to describe a change — Mark rewrites just that card, streams a preview, and swaps in the new version when you apply it (one undo away).
You might ask it to tighten a paragraph or turn prose into a bullet list.
Every tool now has a number-key shortcut alongside its letter, and a new Hand tool (H) lets you drag the canvas around without holding Space.
With the select tool, a single click now picks up a pen stroke, rectangle, ellipse, arrow or icon — you no longer have to drag a box around it first. Shift-click adds or removes one from the current selection.
Pop-up dialogs now sit as one clean, rounded card — the footer bar no longer peeks outside the rounded edge, including in dark mode.
The icon tool now only drops an icon when you actually click on the canvas — panning with the middle mouse button or picking an icon from the toolbar no longer leaves a stray icon behind.
The selection toolbar now labels its Stroke, Fill and Border colours, only shows the stroke colours when they apply, and lets you remove a border entirely with a new Border “None” option.
Two new live demos on the home page show Mark's AI in action: prompting a card onto the canvas, and reshaping an export with an instruction. Shared links now come with a proper preview image, too.
The home page now demonstrates Mark instead of describing it: an animated canvas plays the whole paste-arrange-order-export story, and every section is skimmable at a glance.
Drop ready-made icons — database, server, code, email, person and more — straight onto the canvas to build clearer diagrams. They export with your markdown like any other sketch.
Click the AI button in the toolbar, type a prompt, and pick a model — Mark drafts a Markdown card and drops it on your canvas, ready to edit and export like any other card.
You might ask it to draft a project brief, then rearrange the pieces.
Drop an image onto the board, paste one from your clipboard, or use the new image button in the toolbar. Images become movable, resizable cards and are included when you export.
Text cards now sit transparent on the canvas by default, and you can give any card or shape a background colour and a border colour from the toolbar when it's selected.
You might tint a card to group it with a sketch of the same colour.
You can now pick which frontier AI model formats your export — from Anthropic, OpenAI, or Google — and Mark remembers your choice as the default for next time.
Every canvas card on the dashboard now shows a small preview of the board — cards, sketches, and all — so you can find the right one by shape, not just by name.
The editor now has a button that takes you straight back to your dashboard. No more editing the URL to get home.
You can now erase pen strokes directly with the new eraser tool, or select any group of elements and delete them with one button.
Signup now tells you a confirmation email is on the way and what to do next, instead of silently dropping you at the login page.
The toolbar ignored clicks while a drawing tool was active. Fixed — you can now switch tools at any time.
Write an instruction at export time — "turn this into a PRD," "rewrite as an email" — and get formatted markdown back side by side with the raw export. The instruction and result are saved with the export so you can re-run it later. Free, with a daily usage limit.
Fixed a bug where the canvas could lock up in an endless re-render after certain edits.
Flip into export mode, click the pieces that matter in the order you want them, and export one clean .md file. Markdown cards export as text, never images, and every export is saved so you can get it back.
Sketch around your cards with pen, rectangle, ellipse, and arrow, in a handful of colours.
A dashboard grid of all your thinkpads, a ⌘K switcher to jump between them, and automatic snapshots of your work as you go.
Paste markdown and it lands as a styled, movable card — headings big, bold bold, lists as lists. Select, move, and resize cards anywhere on the canvas.
The first working build: sign up, log in, and get an infinite canvas that pans, zooms, and saves your work automatically while you think.
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.