Shopify Canvas Lets Merchants Build Stores by Chatting With AI

Shopify introduced Canvas on October 1, 2026, a store-building interface that lets merchants assemble an online store by chatting with its Sidekick agent. TechCrunch

Sidekick is the AI assistant inside the Shopify admin. A merchant describes the store, a section, or a style change in natural language, and Canvas applies the edit while the conversation continues. The result appears in real time as Sidekick works.

The output is production code. Canvas renders the store's real code rather than a static preview, according to Shopify. Merchants can test full page interactivity and animation inside Canvas and check pages across different screen sizes. A mockup, a rough visual draft, can hide broken buttons or layouts. Live theme code, the files that control how the store looks and works, does not.

Interaction runs in both directions. Merchants can click on an individual element in Canvas to change it directly, rather than describing its position in text. They can pull back to see the entire store at once, then zoom and pan while updates land in the site's actual files. Sidekick takes screenshots of the Canvas work as it goes, so it sees the same view as the merchant while the site updates.

To enable that loop, Shopify gave Sidekick the ability to work directly with theme files and simplified the theme architecture, the basic structure behind those files. The company did not detail the schema changes, the technical rules for that structure, in the launch materials, but the practical effect is a shorter path from prompt to file write. Sidekick was already used to write code, build apps, customize themes and make other edits. Canvas moves that capability into a shared visual workspace where chat and visuals stay together.

Shopify said Canvas was rolling out over the coming days from its announcement. Shopify

The broader context here is the shift in where storefront work happens. Theme editing has long meant alternating between code, preview, device toolbar and admin settings. Canvas collapses those surfaces into one chat-plus-canvas loop, with visual grounding from screenshots and direct manipulation for corrections. In my view, that combination is the interesting part. Pure prompt-to-site generators are fast but difficult to steer. Direct selection plus conversational instruction gives developers and merchants a tighter correction mechanism, and operating on live files removes a translation step that often introduces drift.

Worth flagging for teams that maintain storefronts at scale, conversational edits still need version control, review and staging discipline. Live code is convenient until a poorly worded prompt touches a shared template. The payoff, if that workflow hygiene holds, is faster iteration on layout, merchandising and responsive behavior without leaving the admin. For small teams without dedicated front-end resources, that lowers the cost of testing ideas in a real storefront rather than in a slide deck.