Skip to content
PLEXON AI
Theme
Download

Design Canvas

Design it whereyou are talking about it

The design canvas is a board of pages docked beside the chat. Ask for a landing page, three versions of a hero, or a mobile and a desktop layout, and they appear as frames rendering live. Click the heading you want changed, say what is wrong, and watch the page update in place. Each frame is an ordinary HTML file in your project, so nothing has to be exported to be real.

Related features
The design canvas filling the window: the Layers tree on the left, a desktop and a mobile frame on the board, the selected heading outlined with its small element toolbar, and the Properties panel on the right

A board of real pages, not a picture of one

Ask for a landing page and it lands on the board as a frame you can read at full size. Each frame is one HTML file under your project, at the size you asked for (desktop 1280x800, mobile 390x844, or anything you name). The pane renders it live: when the AI edits the file, a watcher sees the write and the frame redraws where it sits.

  • Frames live at .plexon/canvas/<board>/frames/ inside your project
  • No separate format and no export step: a frame is a page
  • Drag the background to pan, the wheel to zoom, Fit to frame the whole board
  • Your git history carries the frames like any other file
Three pricing page directions side by side on one board, each a full 1280 by 800 frame, with the Layers panel listing Pricing A, Pricing B and Pricing C

Three directions, side by side

Ask for a few alternatives and they arrive as frames laid out left to right, so you compare them at their real size instead of flipping between tabs. The AI screenshots each frame after it writes it and reads its own work back, which is why the second version usually fixes something you would have had to point out.

  • A new frame is placed beside the rightmost one, never on top of it
  • The AI reviews its own frame with a screenshot before starting the next
  • Pick one and keep working on it; the others stay on the board
  • Send an options set from the chat to the canvas to compare the candidates there

Point at the part you mean

Move the pointer over a frame and the element under it outlines. Click it and a small toolbar appears with its tag name. From there you can leave a comment, hand it to the AI, read its markup, or type straight into the text. The Layers panel shows the same page as a tree, and Properties shows what the element actually computes to, including the values a stylesheet gives it.

  • Layers is a searchable tree; picking a row selects the element on the frame
  • Ctrl-click (Cmd on a Mac) picks several elements, and Properties marks every value they disagree on as Mixed
  • Properties reads live computed values: size, weight, colour, padding, radius, opacity
  • The opacity slider changes the frame while you drag it and undoes as one step
  • The HTML tab shows the element and everything inside it, markup coloured, and takes an edit
  • Double-click text to place the caret and edit in place; the file is saved for you
The Pricing directions board with the Comments panel open on the right, three resolved threads listed under Pricing C, and one of them open on the frame showing the question and the answer

Every comment the board has, in one list

A comment the AI answers and resolves in the same turn sits on the frame for about a second. The Comments panel is where you read it back. It lists every thread under the frame it was left on, and the number on the toolbar button counts the ones still waiting on somebody. Click a row and the board flies to that element and puts the conversation back on the frame.

  • Filter the list to Open, Resolved or All, and narrow it to one frame
  • A resolved thread keeps a muted pin with a check where it was left
  • Reopen one from the list or from the card when the answer did not land
  • Arrow keys walk the rows, Enter opens a thread, N jumps to the next one still waiting
  • Fold a side panel away and it leaves a 28px strip with a chevron to bring it back
The chat on the left with an Open canvas chip under a reply, and the same board on the right showing the desktop and mobile frames

The board and the chat are one conversation

Say "make the hero taller on the desktop one" and the AI edits that file. You do not have to say where anything is. Click an element first and the message carries it: the composer shows what you scoped to, and the AI gets the selector, the markup and a cropped picture of exactly that element.

  • Pin a comment on an element and it waits on the frame until the AI answers it
  • Ask AI sends the comment as a request and resolves the comment when the work lands
  • A rewrite is patched into the page in place, so the scroll position and your selection survive
  • A large rewrite plays back over a few seconds instead of blinking
  • Undo and redo work on the board: Ctrl+Z steps back through moves, resizes and style changes
The design canvas holding two imported pages of a PDF report side by side, a comment pinned on a name at the top of the first page, and the Comments panel listing that thread

Start from a design you already have

Import PDF, in the More menu, puts a PDF on the board one frame per page at the page’s real size; dropping the file on the board does the same, with the first page where you dropped it. The words on a page stay selectable, so a comment pins on a heading and the AI reads the copy. Imported frames are never edited: ask for a rebuild and an editable page lands beside the original, keeping the document’s own typefaces, its logos and pictures, and the position of everything on the page. The import saves those fonts and images beside the frame and writes down where each line, rule and picture sits, so the rebuild is measured rather than guessed from a screenshot.

  • Up to 40 pages per import; pages appear as they render, and Stop is one click
  • The page text sits in a plain file beside the frame, so the AI works from the real copy
  • The page’s geometry sits in a second file: fonts, text lines, rules, bands and images
  • The AI can run the same import from the chat with its canvas tool
  • Delete the frames to take an import back; there is no undo for it
The Import from Figma dialog over the board, with a Figma frame link pasted in and the Import button ready

A Figma frame from its link

From Figma URL, under Add frame, takes the link Figma gives for a selected frame and puts that frame on the board at its own size, drawn by Figma itself. A file link imports the first page’s frames, up to 12. The first time, Plexon asks for a Figma personal access token and holds it until the app closes; Settings keeps it for good.

  • Faithful without local fonts: Figma renders the frame, Plexon adds the selectable text layer
  • The AI also receives the design’s colours, type and layer tree, to rebuild it as an editable page beside the import
  • Prototype links between frames and the frame a flow starts on come in too, and a link to a page imports that page’s frames
  • Token, not an OAuth app: the same token the Figma tool uses, asked for once
  • Imports go into the board’s own folder, so they travel with a backup or a transfer

What you settle stays settled

The Memory panel holds the board’s own house rules, the decisions that came out of your comments, and any frame you pinned as a reference. When the AI reads enough settled decisions to see a pattern, it offers one rule for you to accept or dismiss. Accepting writes it into the board’s guidelines and keeps the previous copy, and the panel lists it.

  • Guidelines are a markdown file you can edit by hand in the panel
  • Every earlier version is kept, and the panel lists them
  • A proposed rule says which decisions it was read from
  • Pin a frame as a reference and the AI matches it instead of guessing

Show it, save it, publish it

Present fills the window with one frame at its own size, and the arrow keys walk the board. The frame stays live while it is up, so an edit the AI makes lands on the screen you are both looking at. When you are done, take the board with you as images, as a PDF, or as a web address.

  • Export PNG or JPG writes the selected frame at twice its size; Copy image puts it on the clipboard
  • Export as PDF writes one file with a page per frame, each page the frame’s own size
  • Publish as gallery puts the whole board on a web address: one page listing every frame at the size you designed it
  • Publishing needs company hosting on your account and a board inside a project
The Build an app from this board dialog over the design canvas: the app name Demo Bakery, the Web target ticked and always built, Mobile and Desktop unticked, the project folder line, and the Build button. Two frames of a bakery site sit on the board behind it.

From board to app

Build app, in the toolbar, turns the board into a working application. Name it, tick Mobile or Desktop if you want them (the web app is always built), and a chat opens beside the board. The AI reads every frame, the house rules and the open comments, writes a brief with the screens, the navigation and the data, and shows it to you as one form. Nothing is built until you confirm it. Then it builds a React web app one screen at a time, checking each one against a screenshot of its frame, and adds the phone and desktop shells you asked for. A journey map or a storyboard works too: each stage becomes a screen through the brief.

  • One web codebase: Android and iOS are a Capacitor shell around it, the desktop build an Electron one
  • Installers are built on your computer when the SDKs are there; otherwise the project is handed over ready to build
  • An apps chip on the board lists every app with its state, plus Open preview, Open chat, Update from board, Publish, Show in folder and Remove from board
  • Update from board rebuilds only the screens whose frames changed
  • Publishing happens only when you ask for it, through the same hosting as your other sites
  • From chat, @skill:canvas-app-builder with the board’s id does the same

It stays on your computer

A board is a folder in your project. Nothing about a canvas reaches the Plexon server, and a publish uploads only the frames and the page that lists them. Your comments, the decisions and the board’s house rules stay on your machine.

  • One folder per board, holding canvas.json and one HTML file per frame
  • Show in folder opens it; delete removes the folder and its frames, with no undo
  • A publish excludes canvas.json, the guidelines, their history and pinned references
  • A board travels in a transfer bundle and in cloud backup, like a dashboard or a data source

A frame that moves is a video

Ask for an animated product card, a reel, or a launch video and the AI writes a frame that says how long it runs and at what frame rate, then animates it with ordinary CSS. The board plays it live, its title bar carries a chip such as Motion · 20 s · 30 fps, and its right-click menu gains Replay, Export as MP4 and Export as WebM. Pick a file name and Plexon records the frame one instant at a time on your computer, so a slow laptop produces the same file as a fast one.

Sound goes in the file too. The frame names its music, sound effects and narration with a start time each, kept in the frame's own folder, and the export lays them under the picture: the music fades at the end and drops under any voice. On the board the frame plays silently; the sound is in the export. Name a moment and it is saved as a poster beside the video, the frame every player shows first. A frame that fetches a font or a script from the web, or animates with something the recorder cannot step, is refused in a sentence before anything is recorded, and the AI fixes it on the board.

This is how brag works. Type it inside a project and the bundled skill storyboards a 15 to 25 second launch video from the project's own code, composes it as one frame on a board named after the project, reviews each scene from a still frozen at the moment it settles, and exports the mp4 with a poster and a share caption. The frame stays on the board, where you can replay it, ask for a change, and export it again. The storyboard path, for a narrated cut of footage you already have, is on the Video Composition page.

What the design canvas gives you

Frames that render live

Each frame is one HTML file in your project. The AI writes it, the watcher sees it, the frame redraws.

Click the thing you mean

Hover to outline, click to select. The message you send carries the selector, the markup and a crop of that element.

Layers and properties

A searchable tree of the page and a panel of its computed values, including what a stylesheet gives it.

Comments on the frame

Pin a comment where the problem is. It stays there until the AI answers it and resolves it.

Versions side by side

Three directions at full size on one board, each reviewed by the AI with a screenshot before the next one starts.

House rules the AI follows

Guidelines, decisions and pinned references live with the board, so a later chat starts from what you agreed.

Bring a piece of any page over

Pick elements in the browser and send them to a board as editable frames, with the page styles written in. Restyle them, then ask for the change to go back into the site.

Documents and code, not only designs

A plan or a source file goes on the board as a page you can edit. Code blocks get their own editor, and the frame exports as Markdown or as source files.

Every frame keeps its history

See the states a frame has been in, what changed between two of them, and put an older one back. Right-click the frame, History.

Present and export

One frame full screen with the arrow keys, or PNG, JPG and a PDF with a page per frame.

Publish the board as a gallery

One web address listing every frame at its designed size. Needs company hosting and a board inside a project.

Build an app from the board

A brief you confirm in one form, then a web app built screen by screen against the frames, with Android, iOS and desktop shells when you tick them. Preview, update and publish it from the board.

A frame that moves

Declare a length and a frame rate, animate with CSS, and the frame is a video: Replay on the board, Export as MP4 or WebM from its menu, sound included.

Import a PDF or a Figma frame

A PDF becomes one frame per page with selectable words; a Figma link becomes a frame at its own size, drawn by Figma.

Files you already own

Frames sit in your repository and your git history carries them. Nothing about a board reaches the Plexon server.

Connects to

Nothing here works alone. See the whole map for how the pieces fit together.

What feeds it

  • AI Chat & Modes authors this.

    Ask for a page, three versions of a hero, or a mobile and a desktop layout, and each one lands on the board as a frame you can read at full size.

  • Dynamic UI Widgets feeds this.

    Open on canvas takes the page a widget is already showing and adds it to a board; Compare on canvas lays every candidate out side by side.

Where it goes

  • This feeds Video Composition.

    A frame that moves exports as MP4 or WebM from its own menu, sound included; brag builds its launch video as one.

  • This renders into Publish Sites.

    Publish as gallery puts the whole board on a web address as one page listing every frame at the size you designed it, and an app built from the board publishes its web build the same way.