updated Claude – Anthropic says its Claude Design—launched in April as an AI tool for “vibe coding” and prototype-making—has received a major update aimed at designers’ biggest complaints. The company claims the new version better follows design systems, offers more precise ed
Designers who tried Claude Design in the months after its April launch didn’t complain about novelty—they complained about control.
The tool. created by Anthropic Lab. is meant to help people move quickly from early ideas to workable prototypes in a “qualitative. vibe-y” space. After it reached designers. Anthropic says it took aim at the pain points those teams reported most: inconsistency with design systems and limited ways to steer what the AI produces.
In the update, Anthropic says Claude Design is now better at adhering to design systems. It also introduces more finely-tuned editing controls, alongside a new efficiency push intended to let users get more done with fewer tokens.
Nate Parrott. an Anthropic designer. said the earlier version of Claude Design didn’t apply design systems with enough consistency across generated prototypes. With the new version. he said. non-designers using the tool can create prototypes that better align with brand and style guidelines. while design admins can gain more ability to control the output.
“ We’ve been continuously hill-climbing our ability for Claude to adhere in the sort of qualitative, vibe-y ways that real designers of real companies want that stuff to happen,” Parrott said.
That shift matters because it changes who can steer early prototypes. When brand and style guidelines aren’t reliably reflected. teams often end up spending their momentum cleaning up what the AI produced. Parrott’s pitch is that the update brings Claude Design closer to the expectations of designers—without forcing teams to wait until late-stage iteration.
The new controls focus on making edits feel more direct. Part of the update includes improved editing tools that allow finer control over layout, type choice, and button styles within an interactive prototype.
“You can go in and you can edit things directly and try your hand at what it looks like with a different font or a different color, and you can get some of those direct controls that you might have had in other tools that designers are familiar with,” Parrott said.
There’s also a practical constraint behind the update: token limits.
To make Claude Design more efficient. Anthropic says Claude Design now shares usage limits with chat. Claude Cowork. and Claude Code. Parrott said the engineering work behind this was driven by demand—users hitting Claude’s token constraints and needing the design tool to do more within the same allowance.
“People can’t get enough Claude tokens, and so we got a lot of work on the engineering side to make Claude Design do more with the same amount of tokens,” Parrott said.
For Anthropic, the goal isn’t just better prototypes. It’s to move Claude Design to the moments when designers are still exploring.
Parrott said the aim is for Claude Design to become a go-to tool when teams are brainstorming and testing new ideas—more about shaping early direction than polishing the final output.
“It’s much more about how can we stake our claim to the beginning of the design process, rather than the end,” he said.
What that promise looks like, internally, is a tug-of-war between volume and judgment. Parrott said Anthropic’s own teams now spend more time sorting through a flood of ideas and directions—deciding which ones should earn engineering attention. which deserve roadmap placement. and which should be explored deeper with more designer time.
“A lot of these AI tools have made it way easier to go from conviction to production,” Parrott said. “What we’ve seen internally is that so much more of our time now is spent on. like. we have a million ideas and a million directions we can go in. Which ones are the right ones?. How do we build conviction?”.
The broader trend is that teams are using AI design tools as “pre-conviction” systems—fast ways to iterate before committing real resources. Parrott pointed to Disney Imagineering. which developed its own bespoke AI tool with Adobe to iterate designs for Disney parks and cruises. He said the value comes from making quick first passes on ideas.
“Previously, somebody would come to me with an idea that was half-sketched out or on paper,” Parrott said. With AI design tools, he said, “we can even decide if it’s worth bringing to engineering or putting on the road map or exploring in depth with more of a designer’s time.”
For now, the market is still searching for how design stacks will work in an AI-forward era—who codes, who approves, and where the handoffs happen. Parrott said he’s heard that widely both inside and outside Anthropic.
“ The big thing he’s heard both internally and externally is ‘that everybody’s trying to figure out what the stack for designers looks like in this era. ’” he said. With more designers empowered to code now. Anthropic is betting that its improved Claude Design—especially the stronger design-system adherence and tighter editing controls—can earn a bigger place in that workflow.
Anthropic Claude Design AI design tool design systems interactive prototypes token limits Claude Cowork Claude Code vibe coding Nate Parrott Disney Imagineering