Canva Magic Layers now works from Claude, ChatGPT, Gemini and Copilot
By Steven Van ·
Images generated in an AI assistant can now be split into editable layers via Canva's MCP Server, without exporting first.
Canva has taken its most powerful AI feature and dropped it right where you are already working. As of July 1, 2026, Canva Magic Layers now pairs with AI assistants — whether you come from Gemini, Claude, ChatGPT, or Microsoft Copilot, you can unlock every visual element of an AI-generated image as a separate, editable layer to remix, refine, and make your own.
From flat AI images to fully editable designs
Magic Layers, first unveiled in March 2026 as a capability of the Canva Design Model, turns any flat image into a structured, multi-layered design inside the Canva editor. Instead of being stuck with a single baked-in picture, you get objects you can reposition, resize, or animate, fonts you can swap, colors you can adjust, and layouts you can rebalance — all in the same space where your work already lives. The feature caught on fast, with Canva reporting it was used more than nine million times in its first four weeks.
Why the AI assistant pairing matters
The new twist is that you no longer have to start inside Canva at all. Generate an image with Claude, ChatGPT, Gemini, or Copilot, and Magic Layers can pull it apart into editable pieces without the usual export-and-rebuild dance. This closes the biggest gap in the AI-image workflow: creative output that looks great but is frozen the moment it is generated. Now the handoff from prompt to polished, on-brand design happens in a single, continuous flow.
Powered by Canva's MCP Server and AI Connector
Under the hood, the integration runs on Canva's Model Context Protocol (MCP) Server, which powers the Canva AI Connector. Once you link your Canva account to your assistant, you can type @Canva to create designs with Canva AI, autofill templates, find existing designs, resize them, and export as PDFs or images — all through natural language. The connection stays private and securely in sync, so brand kits and assets carry through. The AI Connector currently works inside Anthropic's Claude, OpenAI's ChatGPT, Microsoft Copilot, and Google's Gemini.
What creators should try first
If you lean on AI images for social posts, thumbnails, or ad variations, the quickest win is generating a concept in your assistant of choice, then invoking Magic Layers to split it into layers and swap headlines, logos, or product shots. For teams, the payoff is consistency: the same shared brand kit that governs your Canva workspace now applies to assets that originated in an entirely different tool. It is a meaningful step toward treating AI generation and hands-on design as one workflow rather than two disconnected steps.
Sources: Canva Newsroom, Introducing Magic Layers, PetaPixel.