How to use Claude Design now that it lives inside Claude

If you are wondering how to use Claude Design today, the short answer is that you no longer open a separate app. Since 16 September 2026, according to Anthropic's Claude release notes, Claude Design works inside any Claude conversation, including Claude Code and the Artifacts tab, "with all of its features, including on-canvas editing and importing your design system." You ask for a screen, a prototype or a one-pager in chat, Claude builds it on a canvas beside the conversation, and you refine it with chat, inline comments or direct edits.

The practical workflow is four moves: import your design system once, brief Claude like you would brief a junior designer, refine on the canvas with the right editing mode for each change, then export or hand the work to Claude Code. The rest of this guide covers each move, where the tool breaks, and what it means for the people whose job is design.

What Anthropic actually shipped

Claude Design first launched as an Anthropic Labs product on 17 April 2026. In June, Anthropic's post on keeping Claude Design on brand added design system imports, an editor for working directly on the canvas, a /design-sync command and a list of export connectors. The September release folded all of it into regular Claude, alongside Claude Slides and Claude Docs.

Kept to what Anthropic states in its Get started with Claude Design help article:

Three ways in. Start a design in a normal conversation, pick a Design template from the Artifacts tab, or type /design in Claude Code. The iOS and Android apps support it too.

Design systems as inputs. You can bring in one or several design systems from a GitHub repo, design files, uploads or a local codebase, and Claude checks its output against your system before you see it.

Three editing modes. Chat for structural changes, inline comments for targeted fixes, and direct editing to drag, resize and align elements.

Real exports. ZIP, PDF, PPTX, Google Slides or standalone HTML, plus connectors such as Canva, Gamma, v0, Vercel and Miro, and a handoff to Claude Code.

Availability. Anthropic lists Free, Pro, Max, Team and Enterprise plans. On Enterprise it turns on by default on 15 October 2026. The standalone site at claude.ai/design closes on 14 December 2026, so move any design systems over before then.

Shared usage. Design draws from the same usage pool as the rest of Claude, including Claude Code. There is no separate design allowance, and big projects with many iterations cost more.

Hype vs reality

The hype says designers can now skip Figma and go straight from a prompt to a shipped product. Some of that is true. A conversation can now produce a clickable prototype, a deck and the code handoff without leaving one window, and a well imported design system keeps the output closer to your brand than a generic generator would.

What does not hold up is the idea that the tool replaces a design process. Anthropic's own help article lists the gaps: no version history yet, unreliable simultaneous editing by several people, comments that can vanish before Claude reads them, lag on very large repositories, and weaker output when the source design system is messy or incomplete. That last point is the important one. Claude Design amplifies the quality of the system you feed it. A team with clean tokens and documented components gets leverage. A team with a half finished library gets fast inconsistency.

Verdict: Claude Design inside Claude is the most convenient way right now to turn an idea into something clickable and shareable, and worth a serious trial on exploration and early prototypes. It is overhyped if anyone treats it as a replacement for a maintained design file, design reviews or usability testing.

Claude Design vs a canvas-first design tool

Pick Claude Design when speed of exploration matters more than precision, or when the person asking has no design tool open. Keep a canvas-first tool as the source of truth for anything a team maintains over months. If you work in Figma already, the Figma agent guide for designers covers the equivalent AI workflow on that side.

The System, Sketch, Ship loop

A practical routine for using Claude Design on real work. These are suggested steps based on Anthropic's documentation, not test results.

System. Before your first prompt, import the design system. Point Claude at the repo that holds your tokens and components, or run /design-sync from Claude Code if your system lives in code. Clean it first: name tokens consistently, delete dead components, and write short usage notes for anything ambiguous. The thinking behind this is in design systems for AI: a system written for machines needs explicit rules, not tribal knowledge.

Sketch. Brief Claude the way you would brief a colleague. State the goal, the audience, the content you actually have and the layout you expect. Attach screenshots of the current product or reference images. Ask for two or three directions instead of one, then use chat for big structural moves ("move pricing above the FAQ", "switch to a two column layout"). Switch to inline comments for component level fixes and to direct editing for spacing and alignment, since typing "move this 8 pixels left" wastes usage.

Ship. Decide what the output is for before you export. A stakeholder review wants a share link or PDF. A deck wants PPTX or Google Slides. A build wants the Claude Code handoff, which keeps the design and the code in one thread. Because there is no version history yet, export a copy at every milestone you might want to return to, and paste important feedback into chat rather than relying on comments alone.

One habit worth adopting: end each session by asking Claude to list which components and tokens it used and where it deviated from the system. That turns a fast draft into something a reviewer can check in two minutes.

Where design work is going

Claude Design moving into every conversation is a signal about the next one to three years. Design is becoming a mode of a general assistant rather than a separate app you open. Product managers, founders and engineers will produce their own first drafts of screens, and those drafts will arrive at a designer's desk already clickable.

That changes where design value sits. The designer becomes the owner of the system the drafts are built from, the reviewer who decides what is good enough, and the person who runs the research the drafts never include. Teams that invest in a clean, documented design system will get coherent output from everyone. Teams that do not will get a flood of plausible screens that disagree with each other. The design engineer role, sitting between system, code and review, gets more important, not less.

The fear: will Claude Design replace UI designers?

It will replace some tasks, honestly. Low stakes landing pages, internal tools, first drafts of decks and quick prototypes for a meeting are now something a non-designer can produce in an afternoon. If most of your week is that kind of production, your role is changing.

What Claude Design cannot do is decide what the product should be, notice that a flow confuses real users, defend accessibility when a deadline pushes against it, or keep a design system healthy over years. Its own output depends on a system someone designed and maintains. The longer version of this argument is in will AI replace UX designers.

Concrete moves for this month:

1. Import your design system and generate the same screen three times. Note every place it breaks a rule.

2. Fix those gaps in the system, not in the prompt.

3. Run one real exploration in Claude Design and one in your usual tool, and compare time and quality honestly.

4. Learn the Claude Code handoff, because the designers who can steer code will set the standard for everyone else.

When Claude Design is the wrong tool

Skip it for a source of truth file that a team edits together, for work that needs reliable version history, and for detailed component design where every state matters. Do not ship its output without an accessibility check and a real content pass, and remember that its usage comes out of the same pool your engineers use for Claude Code.

For teams that want AI-assisted design without losing craft, DIGCY offers UI/UX design services and AI product development. DIGCY is an award-winning product design agency in Casablanca and a Dribbble Selected Agency that works with clients globally.

FAQ

What is Claude Design?

Claude Design is Anthropic's AI design tool for creating designs, prototypes, slides and one-pagers. Since 16 September 2026 it works inside any Claude conversation, Claude Code and the Artifacts tab, with on-canvas editing and design system import.

How do I start using Claude Design inside Claude?

Ask for a design in any Claude conversation, choose a Design template from the Artifacts tab, or type /design in Claude Code. Claude builds the design on a canvas beside the chat, and you refine it with chat, inline comments or direct editing.

Is Claude Design free?

Anthropic lists Claude Design on every plan, including Free, Pro, Max, Team and Enterprise. It draws from the same usage limits as the rest of Claude, including Claude Code, so large projects use up your allowance faster.

Will Claude Design replace UI designers?

No, but it will absorb routine production such as quick prototypes and simple landing pages. It cannot set product direction, run research or maintain a design system. Designers who own the system and the review process gain leverage from it.

Is Claude Design overhyped?

Partly. It is genuinely fast for exploration and handoff to code, but Anthropic itself lists missing version history, unreliable multi-person editing and weaker output from messy design systems. Treat it as a strong drafting tool, not a full design process.

What happens to the standalone Claude Design site?

Anthropic says the standalone site at claude.ai/design closes on 14 December 2026. Move your design systems into Claude before then so they stay available in your conversations.