Anthropic has released a major update to Claude Design, turning the product from a visually impressive preview into a more practical tool for teams that need brand consistency, code integration, and better usage efficiency.
The redesign comes about two months after Claude Design launched as a research preview and quickly drew more than a million users in its first week. That early popularity also exposed a problem. Users reported that the product could consume large amounts of tokens very quickly, making it difficult for subscribers on lower-tier plans to use for more than short sessions. Anthropic says the new version addresses that issue while expanding the product's role inside the company's wider AI ecosystem.
The biggest change is a rebuilt design system import workflow. Users can now bring design systems into Claude Design from a GitHub repository, design files, or uploaded assets. Once imported, Claude can generate work using those components and check its output against the approved system before showing results.
For larger organizations, Anthropic has also added an admin role that can approve a single system and restrict edits. That setup is meant to help companies keep outputs aligned with internal brand rules, from typography and color use to spacing and component behavior.
The update marks a shift from Claude Design's earlier positioning as a prompt-driven creative canvas. In that version, the tool could create polished mockups, but the results were not necessarily tied to a company's own design language. The new import and validation process is meant to make the product more useful for enterprises that need consistency across many teams and assets.
Anthropic is also tightening the connection between Claude Design and Claude Code. Developers can use a /design-sync command inside Claude Code to pull a local codebase's design system into Claude Design, helping prototypes start from the same components used in production.
Once a design is ready, Claude Code can pick up the work without requiring a screenshot or manual rebuild. The integration also works in reverse, with a /design command in Claude Code allowing developers to create, edit, and sync design projects directly from the terminal.
Anthropic is betting that this shared workflow can reduce one of software development's most persistent pain points: the handoff between design and engineering. By keeping the same AI system involved on both sides, the company is trying to narrow the gap between what gets mocked up and what gets shipped.
The company says it has also reduced average token consumption per turn and cut error rates. Another change is that Claude Design now shares usage limits with Claude chat, Claude Cowork, and Claude Code rather than drawing from a separate pool. That should give many users more headroom than before.
A new drag-and-resize editor is designed to reduce the need for repeated model turns during small adjustments. Anthropic also says it has made hundreds of stability fixes, which should lower the number of wasted generations caused by errors.
Still, the economics of generative design remain a challenge. Producing complete visual assets requires the model to handle layout, content, spacing, typography, and responsiveness at once, which is inherently more expensive than a simple chat exchange. Anthropic's changes appear to ease that burden, not eliminate it.
Claude Design is also being positioned as a starting point for work that moves into other tools. New export destinations include Adobe, Base44, Canva, Gamma, Lovable, Miro, Replit, Vercel, and Wix, alongside PDF and PowerPoint.
The expansion fits Anthropic's broader strategy of embedding Claude across creative work, coding, and enterprise operations. The company has recently expanded into products for small businesses, financial services, and enterprise workflows, while continuing to link its tools through shared context and connected systems.
Claude Design's overhaul suggests Anthropic wants it to be more than a novelty for generating mockups. The company is aiming to make it part of a larger workflow where design, code, and deployment are handled within a connected AI stack.