ANTHROPIC SHIPS MAJOR CLAUDE DESIGN OVERHAUL AND FINALLY FIXES THE TOKEN-BURNING PROBLEM THAT WAS WRECKING BUDGETS Anthropic pushed a substantial update to Claude Design this week that addresses the product’s most persistent complaint head-on: it has been consuming tokens at a rate that made large-scale design work economically unviable for many teams. The new version repositions the product as a design system compliance layer rather than a freeform generation tool. The overhaul ships with three major changes. Design system imports let teams bring their own component libraries and brand tokens directly into the workflow, so the model is constrained from the start to produce only what the system allows. Code round-trips let engineers take AI-generated design output and push it back into the codebase without manual translation. And the token fix applies more aggressive context pruning to eliminate the redundant processing that was inflating costs on complex multi-screen workflows. For enterprise design teams, the design system import feature is the most significant change. It means Claude Design can act as a guardrail rather than a generator that needs correcting after the fact. If your design system says buttons are blue and 12 pixels tall, Claude Design cannot produce red ones. Anthropic is explicitly targeting the workflow gap between designers who use visual tools and engineers who need deployable code. Whether the token fix is enough to make the economics work at scale will depend on how hard teams push it, but the direction is clearly toward a more disciplined and cost-predictable product. Keywords: Anthropic Claude Design, AI design tools, token optimization, enterprise design system