While design tools compete on "one prompt, one design," Replit's new release goes the other direction: instead of pushing you to write prompts, it hands you a stack of "suggested next steps" cards.
On August 4 (the original blog post was July 29), Replit upgraded its Canvas into Replit Design, with Design and Build tabs side by side in the project header—no export needed within the same project. Select a generated design frame and a "Suggested next steps" card appears; each suggestion is one sentence, and clicking it generates a new frame while the original stays untouched. It's a fork, not an overwrite. The official documentation lists three categories of suggestions: change the look, add a missing piece, or change context (e.g., a mobile viewport).
Correcting a Misreading: Not "Prompt-Free"
This is not a "no-prompt" tool. The official documentation's first sentence is "generate design frames from prompts"—the first frame still requires one of four starting points: a prompt, a template, an import (Figma / URL / screenshot / Claude), or a design system. Suggestion cards are a second step on top of existing output, derived only from the frame you select, not from your behavior or environment. Per the docs, the feature solves "facing a blank prompt box and not knowing what to ask," not "never having to speak." Chat and suggestion cards work alongside each other: "Chat isn't just for polish."
The Real Difference: No Handoff
The difference isn't "being good at design"—it's "no handoff." With v0 or Figma Make, you still need to move your finished design somewhere else; in Replit Design, clicking Build turns the design into a running app within the same project, or conversely reskins an existing app. Cursor and Claude Code are code-side agents lacking a canvas and visual forking layer, so they aren't direct competitors—Claude actually opens a "Send to Replit" channel, pointing toward Claude handing work off to Replit.
What's genuinely new is making forking + side-by-side comparison the default. When you go down the wrong path, you don't undo—two versions sit side by side and you pick. This fits non-designers' cognitive strengths: people are good at choosing between two options, not at describing something from scratch. Design systems land as tokens.json + CSS variables + DESIGN.md, readable by both humans and models—more actionable than a brand-guidelines PDF.
Caveats
- Mock frames on the canvas look like real pages but store no data and connect to no external services.
- Applying a design system to an existing app genuinely regenerates the entire styling layer; Replit shows a time/cost warning first, with no published quota.
- Mobbin integration offers 600,000+ real interfaces across 1,000+ apps (per Replit's own figures), but inspiration never lands in code: "Inspiration adds no code to your project."
- Starting a new app from a design frame requires Core ($25/month) or Pro ($100/month); the "available to all users today" launch-page phrasing is easy to misread.
- Browser only. No independent third-party testing at scale yet.
- Official blog: https://replit.com/blog/introducing-replit-design
- Official docs (overview): https://docs.replit.com/design/what-is-replit-design
- Official docs (suggestion cards): https://docs.replit.com/design/explore-suggestions
- Third-party verification (billing/model list cross-check): https://best.xiaohu.ai/article/replit-design/