Claude Design is Anthropic's latest creative tool from its Anthropic Labs program — a collaborative design workspace where you describe what you want and Claude builds it on a canvas next to the chat. Think prototypes, slide decks, landing pages, mockups, and microsites, all shaped through conversation. It launched on April 17, 2026, and runs on Claude Opus 4.7, Anthropic's most capable vision model anthropic.
Here's everything you need to know to use it well and manage it for a team.
1. Who can use it, and where to find it
Claude Design is available in research preview to Pro, Max, Team, and Enterprise subscribers anthropic. On Enterprise plans, it's turned off by default and admins have to enable it in Organization settings.
You launch it at claude.ai/design. The interface has two halves: chat on the left, a live canvas on the right.
2. Set up your design system first (this is the big one)
The single biggest lever for output quality is your design system — a reusable bundle of your brand's colors, typography, components, and layout patterns that every project inherits automatically.
Rolling out Claude Design without a design system in place means your team will get functional but generic output. So do this first.
What Claude can ingest to build your system:
A codebase or component library (for example, a React library)
Prototypes, screenshots, or existing design files
Slide decks or PDFs that reflect your brand — Claude extracts colors, layout, and typography from them
Individual assets like logos, color palettes, and typography specimens
The setup flow (from the lower-left project picker, click your org name, then complete onboarding):
Create or switch to your organization.
Upload assets.
Review what Claude extracts — typically a color palette, typography, components like buttons/cards/nav, and layout patterns such as spacing and grids.
Validate with a test project ("create a landing page for [product]", "design a dashboard showing [metrics]").
Flip the Published toggle on. After publishing, new projects created while in your organization will use your design system instead of the default.
Pro tip from Anthropic: include real examples, not just specs — a finished landing page tells Claude more about your brand's feel than a color palette alone.
You can have more than one design system per org (useful for sub-brands), and you can update them later via the Remix button inside the design system editor.
3. Start a project and prompt like a pro
Every new project auto-inherits your org's design system, so there's zero configuration on each new file.
Add context up front. The more Claude knows, the less it has to guess:
Screenshots of existing designs, competitors, or visual inspiration
A linked code repo so Claude understands your real components and architecture
Existing slides or docs with a style you want to echo
Write a prompt that hits the four essentials:
Goal — what you're building
Layout — how it should be arranged
Content — what information to display
Audience — who will use it
Examples that work well, straight from the docs:
A dashboard showing monthly revenue with filters for region and product line
A mobile app onboarding flow with four screens
A landing page for a new API product with hero, code examples, and pricing
A feedback form with conditional questions based on category
If you leave gaps, Claude will ask clarifying questions rather than guess.
4. Iterate: chat vs. inline comments
This is where the real design happens. The first generation is a starting point, not the finish line.
Use chat for big moves:
Structural changes ("put metrics on top, chart below")
Aesthetic shifts ("make it darker and more minimal")
New sections, alternative layouts, or asking Claude to review for accessibility
Use inline comments for surgical tweaks:
Click directly on an element and leave a targeted request — "make this button padding larger," "change this to a dropdown," "use the primary brand color here"
The rule of thumb: comments for targeted, component-level changes; chat for structural changes, new sections, aesthetic shifts, or anything that requires explanation or context.
You can also ask Claude for variations — "show me 2–3 alternative layouts" is dramatically faster than guessing which direction is right.
Heads up on a known quirk: inline comments occasionally disappear before Claude reads them; paste the feedback directly into chat as a workaround.
5. Branch without losing work
Want to explore a wild new direction without torching your current version? Just tell Claude something like: "Save what we have and try a completely different approach." Claude will save the current state and confirm where, so you can jump back in the conversation history.
6. Export and share
Hit the Export button in the upper right. Your options:
Download as
.zipExport as PDF
Export as PPTX
Send to Canva (via the official partnership — designs become fully editable there)
Export as standalone HTML
Handoff to Claude Code — either to your local coding agent or Claude Code Web
Sharing within your org supports view-only, comment, or edit access via link. Edit access lets teammates modify the design and chat with Claude in a group conversation.
7. Admin & rollout (Team / Enterprise)
If you're administering this for a team, the official recommendation is a four-phase rollout rather than flipping the switch for everyone on day one:
Design system setup — 2–4 trusted designers across brand and product build and validate the system.
Design team onboarding — full design team stress-tests it on real projects.
Product & UX onboarding — PMs, UX researchers, and adjacent functions come in.
Broader organization — everyone else, or specific departments.
Access to design system setup is controlled separately from general Claude Design access via custom roles claude, so you can give trusted designers editing rights without opening the floodgates.
A few things to know as an admin:
Claude Design doesn't currently support audit logs or usage tracking, so adoption monitoring is qualitative right now — check in with groups and sample projects periodically.
Uploaded assets are stored persistently and fall under the same data retention and deletion policies as other Anthropic enterprise products.
Data residency requirements aren't supported yet.
It's only available via the web interface at claude.ai/design — if you need it through a cloud-provider agreement, contact your Anthropic rep.
8. Pricing and usage
Claude Design is metered separately from regular chat and Claude Code — design activity never draws from those other limits.
Individual plans (Pro, Max 5x, Max 20x): recurring weekly allowance that resets every seven days. Extra usage is purchasable.
Team plan: each provisioned user gets a recurring weekly allowance. Standard seats for light use, Premium for power users.
Enterprise (legacy seat-based): same model — weekly allowance per provisioned user.
Enterprise (usage-based): billed at standard API rates under your existing agreement, with a one-time credit per user covering roughly 20 typical prompts that expires July 17th.
Allowances are granted per user, not pooled at the org level — each provisioned user gets their own bucket.
9. Tips for getting great output
Paraphrased from Anthropic's official tips:
Start simple, layer in complexity. Core layout first, then interactions, then polish.
Be specific in feedback. "Tighten spacing between form fields to 8px" beats "this looks off."
Name your components. If your system has a "Primary Button" or "Card layout," reference it by name.
Mention responsiveness early. Mobile? Tablet? Desktop? All three? Say so up front.
Ask for variations. Two or three options side-by-side beat one lucky guess.
Treat Claude as a collaborator. Ask it to review for accessibility, contrast, or hierarchy — not just to generate.
Which plans include Claude Design?
Claude Design is available in research preview to Pro, Max, Team, and Enterprise subscribers. It's included with your plan and draws from your subscription's Claude Design allowance (which is metered separately from chat and Claude Code). On Enterprise plans it's off by default — an admin has to enable it in Organization settings before anyone on the team can access it. If you need more capacity than your weekly allowance provides, extra usage can be purchased on paid plans.
Can we have multiple design systems for different brands or sub-teams?
Yes — an organization can maintain more than one, which is handy if you run multiple products or sub-brands.
What happens if someone starts using Claude Design before the design system is set up?
They'll still get functional output, but it won't reflect your brand. Anthropic strongly recommends finishing design system setup first.
Can I restrict Claude Design to specific departments?
Yes. Use custom roles (RBAC) to grant access to specific groups rather than enabling it org-wide.
Can we export or archive generated designs?
Yes. Current export paths: HTML bundles, PPTX, PDF, Canva, .zip download, and handoff to Claude Code. More formats are on the roadmap — contact your sales rep if you need something specific.
Can admins see who's using what?
Not yet. Claude Design doesn't currently support audit logs or usage tracking. Adoption monitoring is qualitative for now — check in with each group as they onboard and sample a few projects.
Does Claude Design support data residency?
Not at this time.
Can I update the design system later when our brand evolves?
Yes. Open it from org settings, hit Remix in the upper right, and iterate with Claude via the chat panel.
Do all team members need to upload brand assets?
No. Once a designer sets up your org's design system, every project created inside that org automatically uses it. Everyone else just starts creating.
