Table of Contents


The Editor class is the main way of controlling tldraw's editor. You can use it to manage the editor's internal state, make changes to the document, or respond to changes that have occurred.

By design, the Editor's surface area is very large. Almost everything is available through it. Need to create some shapes? Use editor.createShapes(). Need to delete them? Use editor.deleteShapes(). Need a sorted array of every shape on the current page? Use editor.sortedShapesArray.

This page gives a broad idea of how the Editor class is organized and some of the architectural concepts involved. The full reference is available in the Editor API.


The editor holds the raw state of the document in its store property. Data is kept here as a table of JSON serializable records.

For example, the store contains a page record for each page in the current document, as well as an instancePageState record for each page that stores information about the editor's state for that page, and a single instanceState for each editor instance which stores the id of the user's current page.

The editor also exposes many computed values which are derived from other records in the store. For example, editor.selectedIds is a computed property that will return the editor's current selected shape ids for its current page.

You can use these properties directly or you can use them in signals.

import { track, useEditor } from "@tldraw/tldraw"

export const SelectedIdsCount = track(() => {
  const editor = useEditor()
  return (

Changing the state

The Editor class has many methods for updating its state. For example, you can change the current page's selection using editor.setSelectedIds. You can also use other convenience methods, such as editor.select, editor.deselect, editor.selectAll, or editor.selectNone.

editor.select(myShapeId, myOtherShapeId)
editor.selectedIds // [myShapeId, myOtherShapeId]

Each change to the state happens within a transaction. You can batch changes into a single transaction using the editor.batch method. It's a good idea to batch wherever possible, as this reduces the overhead for persisting or distributing those changes.

Listening for changes

You can subscribe to changes using editor.store.listen. Each time a transaction completes, the editor will call the callback with a history entry. This entry contains information about the records that were added, changed, or deleted, as well as whether the change was caused by the user or from a remote change.

editor.store.listen(entry => {
  entry // { changes, source }

Remote changes

By default, changes to the editor's store are assumed to have come from the editor itself. You can use editor.store.mergeRemoteChanges to make changes in the store that will be emitted via store.listen with the source property as 'remote'.

If you're setting up some kind of multiplayer backend, you would want to send only the 'user' changes to the server and merge the changes from the server using editor.store.mergeRemoteChanges. (We'll have more information about this soon.)

Undo and redo

The history stack in tldraw contains two types of data: "marks" and "commands". Commands have their own undo and redo methods that describe how the state should change when the command is undone or redone.

You can call editor.mark(id) to add a mark to the history stack with the given id.

When you call editor.undo(), the editor will undo each command until it finds either a mark or the start of the stack. When you call editor.redo(), the editor will redo each command until it finds either a mark or the end of the stack.

// A
editor.mark("duplicate everything")
// B

editor.undo() // will return to A
editor.redo() // will return to B

You can call editor.bail() to undo and delete all commands in the stack until the first mark.

// A
editor.mark("duplicate everything")
// B

editor.bail() // will return to A
editor.redo() // will do nothing 

You can use editor.bailToMark(id) to undo and delete all commands and marks until you reach a mark with the given id.

// A
// B
// C

editor.bailToMark("first") // will to A

Events and Tools

The Editor class receives events from the user interface via its dispatch method. When the Editor receives an event, it is first handled internally to update editor.inputs and other state before, and then sent into to the editor's state chart.

You shouldn't need to use the dispatch method directly, however you may write code in the state chart that responds to these events.

State Chart

The Editor class has a "state chart", or a tree of StateNode instances, that contain the logic for the editor's tools such as the select tool or the draw tool. User interactions such as moving the cursor will produce different changes to the state depending on which nodes are active.

Each node be active or inactive. Each state node may also have zero or more children. When a state is active, and if the state has children, one (and only one) of its children must also be active. When a state node receives an event from its parent, it has the opportunity to handle the event before passing the event to its active child. The node can handle an event in any way: it can ignore the event, update records in the store, or run a transition that changes which states nodes are active.

When a user interaction is sent to the editor via its dispatch method, this event is sent to the editor's root state node (editor.root) and passed then down through the chart's active states until either it reaches a leaf node or until one of those nodes produces a transaction.

A diagram showing an event being sent to the editor and handled in the state chart.


You can get the editor's current "path" of active states via editor.root.path. In the above example, the value would be "root.select.idle".

You can check whether a path is active via editor.isIn, or else check whether multiple paths are active via editor.isInAny.

editor.store.path // 'root.select.idle'

editor.isIn('root.select') // true
editor.isIn('root.select.idle') // true
editor.isIn('root.select.pointing_shape') // false
editor.isInAny('editor.select.idle', 'editor.select.pointing_shape') // true

Note that the paths you pass to isIn or isInAny can be the full path or a partial of the start of the path. For example, if the full path is root.select.idle, then isIn would return true for the paths root, root.select, or root.select.idle.

If all you're interested in is the state below root, there is a convenience property, editor.currentToolId, that can help with the editor's currently selected tool.

import { track, useEditor } from "@tldraw/tldraw"

export const CreatingBubbleToolUi = track(() => {
  const editor = useEditor()

  const isSelected = editor.isIn('root.bubble.creating')

  if (!editor.currentToolId === 'bubble') return
  return (
    <div data-isSelected={isSelected}>Creating Bubble</div>


The editor's inputs object holds information about the user's current input state, including their cursor position (in page space and screen space), which keys are pressed, what their multi-click state is, and whether they are dragging, pointing, pinching, and so on.

Note that the modifier keys include a short delay after being released in order to prevent certain errors when modeling interactions. For example, when a user releases the "Shift" key, editor.inputs.shiftKey will remain true for another 100 milliseconds or so.

This property is stored as regular data. It is not reactive.

Common things to do with the editor

Create shapes

    type: 'geo',
    x: 0,
    y: 0,
    props: {
      geo: 'rectangle',
      w: 100,
      h: 100,
      dash: 'draw',
      color: 'blue',
      size: 'm',

Update shapes

const shape = editor.selectedShapes[0]

    id: shape.id, // required
    type: shape.type, // required
    x: 100,
    y: 100,
    props: {
      w: 200,

Delete shapes

const shape = editor.selectedShapes[0]


Get a shape by its id


Move the camera

editor.setCamera(0, 0, 1)

See the tldraw repository for an example of how to use tldraw's Editor API to control the editor.

Edit this page
Last edited on 22 March 2023