Anthropic ships major Claude Design overhaul with design system imports, code round-trips, and a fix for its token-burning problem

Anthropic quietly shipped one of the more consequential updates to its design tooling this week, and the headline fix isn't glamorous but it's the kind of thing that actually determines whether teams keep using the product. Claude Design was eating tokens at a rate that made heavy use painful, and Anthropic has now rebuilt the underlying process so sessions don't bleed cost the way they used to.

The bigger shift is what this means for how the tool fits into real company workflows. Claude Design can now pull in existing design systems directly from GitHub repos, design files, or straight uploads, which means a team's actual component library, spacing rules, and brand colors become the starting point instead of Claude guessing at a generic layout. That's a meaningful jump from a novelty AI design toy toward something closer to an enterprise platform, one that respects the constraints a design team already has instead of ignoring them.

There's also a code round-trip capability baked in now, letting designs move back and forth between design output and actual code rather than staying stuck as static mockups someone has to manually translate. For teams that have complained about AI tools producing pretty pictures that don't map cleanly to a real codebase, this closes a gap that's been nagging at the whole category of AI design tools since they started showing up.

Anthropic isn't alone in chasing this space, but the combination of cost fix and enterprise import features suggests the company is trying to get ahead of the criticism that generative design tools are expensive toys rather than production-ready infrastructure. Whether it holds up under real agency and product-team workloads is the next test, but the direction is clear: less demo, more tool.