Get Lifetime License (opens in new tab)

Working with the AI

The chat on the right of the workspace is your integrated AI partner. It stays with you in every mode, running locally on your own machine with direct BYOK provider keys (Google AI Studio, Anthropic, or OpenAI) and zero token markup.

Unlike isolated prompt boxes that operate in a vacuum, the AI partner is grounded in your live project context: it reads your active route tree, block content, assets, design tokens, and strategic brief.

Project awareness and iterative design

Website design in Qanvas is a continuous, multi-turn collaboration rather than a one-shot generation prompt:

  • Iterate through conversation: Discuss design directions, explore layout alternatives, refine messaging, and adjust sections across successive conversations without starting over from scratch.
  • Dual-modality control: Use the AI partner for rapid structural composition and copy iteration, and pair it with the point-and-click visual inspector on the canvas for instant adjustments to spacing, typography, and styling—saving AI tokens for high-leverage creative work.
  • Retain creative direction: The AI suggests and stages changes, but you retain full creative authority. Every proposal is checked against machine-validated component schemas before staging, and nothing touches disk until you explicitly save.

What the AI can do

  • Shape your strategy: Ask the questions that fill in your site brief, and record the answers you confirm.
  • Write your specifications: Draft and update sections of your living PRD and TDD specifications.
  • Build the site: Plan pages and compose each route from structured Qanvas blocks once your brief is complete.
  • Refine pages and blocks: Change copy, swap or reorder blocks, adjust layout modes, rename routes, and update global header, footer, and theme tokens.
  • Adjust colours: Derive balanced light or dark palettes from your active mode or brand logo.
  • Prepare a support ticket: Package issues or feature ideas into structured tickets for direct dispatch.

While the AI works, its steps appear in the chat, for example Editing blocks on /pricing or Updating PRD.md › Goals. When it changes something, Qanvas switches to the mode where you can see the change.

Sending a message

Type in the box at the bottom (Type a message, instruction, or drop an image…) and press Enter.

  • Attach a screenshot or file with the attachment button, or drop an image into the chat.
  • Paste long text and it is attached as a collapsible snippet.
  • Choose a model from the picker in the composer. It lists the models of every provider whose key you have saved.
  • Stop a reply at any time with Stop generating.

Good requests name what to change and where, for example "Make the hero headline on the home page shorter and more direct" or "Add an FAQ section to /pricing with five questions about billing".

How AI changes are applied

The AI never writes files directly. Every change it makes is checked against the rules of each block and staged into your unsaved draft:

  • Valid changes appear on the artboard straight away, with a notification that the AI's changes were applied (unsaved).
  • Invalid changes are rejected before anything reaches your draft, and the AI retries.
  • One Undo reverts a whole AI change.
  • Nothing is written to disk until you select Save.

Undo, redo and the mode switcher are locked while the AI is working, so a change in progress is never interrupted.

Threads and history

Conversations are kept as threads in your Qanvas data folder.

  • New thread starts a fresh conversation.
  • Open the thread list to switch, Rename or Delete a thread.
  • Search across every thread and message, or use Find in Thread to search the current conversation.
  • Bookmark a message to find it again in the Bookmarks drawer.
  • Edit one of your messages and select Save & Resend to try again from that point.
  • Copy a message or a code block.

In long conversations, earlier messages are summarised automatically so the AI keeps its context.

Retention

Chat retention settings let you delete old threads automatically by age, cap how many threads are kept, and protect threads that contain bookmarked messages from automatic deletion.

Workspace memory

As you work, the AI learns lasting facts about your project, such as decisions, preferences and domain terms. Open Workspace memory to review them, Forget this fact to remove one, or Reset workspace memory to start over.

When something goes wrong

If a reply fails, the chat explains why, for example Rate Limited, Authentication Failed, Connection Timed Out, Model Temporarily Overloaded or Spending Cap Exceeded, and offers Retry.

  • Authentication Failed: check your key in Settings → AI Engine.
  • Spending Cap Exceeded: raise your spend cap with your provider. For Gemini, Manage Spend Cap opens Google AI Studio.
  • No AI provider keys configured: add a key first. See Add your AI key.

Reporting a problem through the chat

Describe a bug or a feature idea and ask the AI to send it to Qanvas support. It prepares a ticket (a bug report, feature request, question or complaint) and shows it to you first. Nothing is sent until you select Confirm dispatch.

Tickets include the name and email from Settings → License → Customer profile, so fill those in first. See Help and feedback.