What is built

Every feature, measured against tldraw: built, partly built or missing.

What Klementine Draw can do, measured against the feature set of tldraw (its SDK's default shapes, tools and documented features, as of its repository at commit 017ae8f, 2026-11-10). Checked on 2026-11-11. This is the one list of what is built and what is not; update it with the code.

tldraw is a trademark of tldraw Inc. Klementine Draw is an independent project, not affiliated with or endorsed by tldraw, and contains none of its code. The comparison is here because tldraw is the most complete whiteboard SDK there is, which makes it the honest yardstick.

Yes: built. Partly: the core of it works, the rest of the line says what is missing. No: not built. Yes does not mean proven in daily use: the last section says how each part was checked, and nobody has worked on a board with this for a day yet.

Shapes

Feature Draw What is there, what is missing
Freehand drawing Yes Pen pressure, hand-drawn stroke (perfect-freehand).
Highlighter Yes Key M.
Geometric shapes (20 kinds) Yes 19 kinds (all but tldraw's second rhombus). Eight are in the toolbar; the others are made from code (geoTool('octagon')).
Text inside shapes Partly Double click and type; left, centre, right and top, middle, bottom. Plain text; the shape does not grow with it.
Text Yes Key T. As wide as its words, or wrapping once resized. Resizing does not scale the type.
Arrows Partly Straight, bent or right-angled, attached to shapes, with words, a head at either end. Missing: a chosen point on the shape, other head styles.
Lines Partly Two ends and a bend. Missing: more than two points.
Sticky notes Partly In the package now (key N), text shrinks to fit. Missing: the buttons that add the next note beside one.
Frames Yes Hold shapes, cut them off, have a name, sit in other frames. Missing: fit to content.
Groups Yes Ctrl G, Ctrl Shift G; double click steps in.
Images Partly Drop, paste or pick a file. Missing: crop, animated pictures are shown as they are, no video files.
Bookmarks and embeds Partly A pasted web address becomes a link card; YouTube and Vimeo addresses a player. Missing: page titles and previews, tldraw's other embed kinds.

Tools and editing

Feature Draw What is there, what is missing
Select, move, resize, marquee Yes Shift keeps proportions or one axis, Alt resizes from the middle.
Lasso select Yes Alt-drag from the empty canvas.
Hand, zoom tool Yes Keys H and Z.
Eraser Yes Key E.
Laser pointer Yes Key K. Other tabs see it.
Rotate Partly Handle above the selection, Shift for 15° steps. Several shapes turned differently cannot be resized together.
Snapping Partly Edges and middles of other shapes; the grid, when switched on in the right-click menu. Missing: equal gaps, a line's ends.
Undo, redo Yes Only your own changes.
Copy, cut, paste Yes Shapes (with their pictures) between canvases; pasted text, pictures and web addresses; copy as a picture in the menu.
Duplicate Yes Ctrl D, or Alt while dragging.
Align, space evenly, flip Yes In the right-click menu. Missing: stretch, stack.
Order, lock Yes
Typing into shapes Partly Plain text. tldraw has rich text (bold, lists, links).
Crop No
Scrolling at the edge Yes A drag that reaches the canvas's edge moves the view.

Styles

Feature Draw What is there, what is missing
Colour (13) Yes 12. No orange, on purpose.
Fill (none, semi, solid, pattern) Yes None, light, full, hatched.
Dash (draw, solid, dashed, dotted) Yes
Size (S, M, L, XL) Yes
Font (draw, sans, serif, mono) Yes The app loads the handwriting font (--kd-font-hand).
Text alignment Yes Left, centre, right; in shapes also top, middle, bottom.
Opacity Yes Five steps.

Canvas

Feature Draw What is there, what is missing
Pan, zoom, fit, zoom to selection Yes
Pinch and two-finger pan Partly Built; not tried on a touch screen yet.
Pages Yes The menu in the top left: go to, add, rename, delete. Each page keeps its view.
Minimap Yes <Draw minimap />. Press or drag in it to go there.
Right-click menu Yes
Light and dark Yes
Grid Yes Show and snap, in the right-click menu.
Skipping shapes off screen Yes Only what is in or near the view is drawn.
Read-only mode Yes <Draw readOnly />: the view moves, nothing changes. <Draw focus />: nothing but the canvas.
Links to a place on the canvas Yes <Draw deepLinks /> keeps page and view in the address.
Keyboard shortcuts Partly Tools, clipboard, group, order, nudge. Missing: the rest of tldraw's set, a shortcut list.
Accessibility Partly Controls are named and keyboard-reachable; shapes are not reachable without a pointer.
Languages Partly English, and a labels prop to translate. tldraw ships about fifty.

Data and platform

Feature Draw What is there, what is missing
Reactive store, diffs, history Yes
Own shapes and tools Yes A plugin is a list of classes.
Bindings Partly tldraw keeps them as records of their own, open to plugins. Draw keeps a line's ties in the line (ties, retie on a util).
Storage in the browser Yes <Draw persistenceKey="…" />: IndexedDB, written record by record.
Sync across tabs Yes Same key: tabs of one browser stay in step.
Cursors of others Partly Pointer, selection and laser of the other tabs, with the name from <Draw user={{ name }} />. Missing: following someone, cursor chat.
Collaboration over a server No The seam is a Channel (post and listen); tabs use it today. A Convex channel and storage are not written: they need a deployed backend.
Checking and migrating stored documents Partly A version is saved with every document, bad records are left out on load, migrate is the place for steps. No step exists yet.
Save and load files Yes Page menu: save to a file, open a file.
Export Partly Page menu: save as a picture (PNG of the page or the selection); toSvg, toPng in code. Pictures from other websites may be left out of it.
Assets Yes AssetStore with one method, upload. By default a file stays in the document as a data URL.
Hooks before and after a change Partly After: store.listen. Before: none.
Replacing parts of the UI Partly Compose Canvas, DrawToolbar, StylePanel, PageMenu, ZoomControl, Minimap, CanvasMenu yourself. tldraw has a slot for every part.
Helpers for AI agents No
Licence MIT tldraw's SDK needs a licence key in production. Klementine Draw contains none of its code and is under the MIT licence.

Decisions made while building (2026-11-11)

  1. Files stay in the document by default. An image is kept as a data URL in an asset record, so it is saved, synced and exported with everything else and needs no server. It makes documents large. An app passes assets: { upload } to the editor to keep files elsewhere (Convex file storage); nothing else changes.
  2. Saving is IndexedDB plus a channel between tabs, behind two small interfaces (DocumentStorage, Channel). A Convex version implements the same two.
  3. Pictures are made from the screen (the shapes' own elements inside an SVG), so a picture looks exactly like the canvas, fonts included. The price: it only works in a browser, and a picture loaded from another website can be missing from it.
  4. A text's size is not stored as truth. Every screen measures its own text; the document keeps the last size only for places that cannot measure.
  5. Sticky notes moved into the package; the example no longer has a plugin of its own.
  6. Sizes, colours, fills follow tldraw's sets, without orange.

Known limits of what is built

  • A diff carries whole records, so two people changing the same shape at the same moment overwrite each other's fields. A server sync has to decide this.
  • An undo in one tab can take back something another tab changed in the same shape since.
  • Deleted pictures stay in the document as asset records (so an undo can bring them back); nothing cleans them up yet.
  • No app uses the package yet: it is run by its example and its tests.

How it was checked

Engine tests (packages/draw/tests, 40, no browser) cover the store, undo, tools, every shape kind's geometry, attached, bent and right-angled lines, text, eraser, rotation, groups, frames, clipboard, arranging, snapping, the grid, pages, saving and tabs (with stand-ins for IndexedDB and the channel), cursors, files and assets, read-only, scrolling at the edge.

No engine test, because the code only runs in a browser: the minimap, links to a view, not drawing what is off screen, picture export and copy as a picture, saving to and opening a file, the hatched fill, focus mode, reading a picture file, IndexedDB itself, dropping and pasting files.

Tried by hand in a browser (Chrome, desktop): text tool, text in shapes, right-click menu and Duplicate, rotating by the handle, frames taking in shapes, eraser, a pen stroke kept across a reload, attached, bent and right-angled arrows with words and two heads, all 19 shape kinds, an image added and kept across a reload, a new page and a page rename from the page menu, picture export (compared by eye), two tabs in step with each other's cursor, laser, minimap, links to a view, shapes off screen not drawn, a link card, a hatched fill, text position.

Not tried by hand: real Ctrl+C and Ctrl+V, dropping a file, the file picker, Alt-drag, lasso, grouping, lock, align, snapping guides, the grid, deleting a page from the menu, save and open a file, copy as a picture, players for videos, read-only and focus mode, scrolling at the edge. Never run at all: pinch on a touch screen.