Anthropic just launched Claude Design, turning prompts into interactive prototypes and design assets—raising fresh questions about Figma’s dominance.
Anthropic has officially launched Claude Design, a new AI product that turns plain-language prompts into polished visual work—designs, interactive prototypes, slide decks, and marketing pages.
For many teams. the appeal is obvious: a workflow that starts with an idea and ends with something clickable and presentable. without requiring the user to master a design tool first.. The bigger question is whether Claude Design meaningfully shifts the UI/UX category itself—or simply adds another layer of automation on top of existing software.
Claude Design turns prompts into prototypes, with real editing controls
Claude Design, available now in research preview for paid Claude subscribers, is built around a “conversational” creation loop.. Users describe what they want. Claude generates a first version. and then refinement happens through a mix of chat. element-specific inline comments. direct text editing. and sliders for adjustments like spacing. color. and layout.
A key part of the pitch is that the tool doesn’t treat every project as a blank canvas.. During onboarding. Claude can read a team’s codebase and design files to build a design system—typography. colors. and reusable components—then apply it to later work.. Teams can refine that system over time and even maintain multiple versions.
The import surface is also designed for speed. Users can start from text, upload images and documents, or point Claude at a codebase. There’s even a web capture tool that pulls elements from a live website, so early prototypes look closer to the eventual product rather than generic mockups.
The handoff to coding is the feature Anthropic wants to own
What separates Claude Design from the broader wave of AI design experiments is the handoff mechanism. When a prototype is ready to build, Claude packages it into a handoff bundle that can be passed to Claude Code using a single instruction.
That matters because the current reality in many orgs is fragmented: designers iterate in one tool. developers implement elsewhere. and feedback loops drag across formats. reviews. and rework.. Claude Design aims to collapse that chain—from exploration to prototype to production code—inside Anthropic’s ecosystem.
Still, Anthropic isn’t only betting on its own next step.. Users can share an internal URL. organize outputs as folders. or export to formats including Canva. PDF. PPTX. and standalone HTML.. The practical goal is clear: if Claude Design is meant to expand the number of people who create prototypes. it has to fit into how companies already collaborate.
Why Claude Design pressures the Figma narrative
The launch lands with extra friction. Around the same time, Anthropic’s chief product officer stepped down from the board of Figma. The timing doesn’t prove anything on its own, but it underlines a structural tension between “partnering on AI” and “competing for the design workflow.”
Figma’s recent move—like converting AI-generated code into editable designs inside its product—was effectively a bet that AI would make design more central. not less.. Claude Design complicates that bet by moving the value earlier: it enables non-designers to generate complete. interactive prototypes from natural language.
That expansion is the competitive threat, even if professional designers still prefer Figma for certain craftsmanship or established team practices.. For startups and product teams operating with lean headcounts. the ability to go from prompt to prototype without weeks of design ramp-up can change how roadmaps are planned.
The new model engine: Opus 4.7, built with constraints
Claude Design runs on Claude Opus 4.7. Anthropic’s most capable generally available model. with improvements tied to software engineering. instruction following. and vision.. The company also released its broader “Mythos” line with more advanced cybersecurity capabilities. but positioned it behind vetted access due to risk.
That dual-track approach is unusual in AI, and it shapes the product in two ways. First, it signals Anthropic is engineering not just capability, but guardrails—by intentionally reducing certain cybersecurity risks during training and adding protections that block prohibited or high-risk requests.
Second, it reinforces the idea that Anthropic is building applications at the same time it refines model behavior. For a design tool, that “behavior tuning” matters because design work often pulls in code, assets, and internal context—everything that needs predictable handling.
Enterprise buyers will judge privacy before they judge polish
For organizations evaluating Claude Design, privacy and data handling will likely be the decisive factor.. Anthropic says its system stores a design-system representation it generates rather than the source files themselves. and that when users connect to a local copy of code. it isn’t uploaded or stored on Anthropic servers.
There’s also an added integration path with GitHub connectivity, plus an explicit statement that Anthropic does not train on the connected data. For Enterprise customers, Claude Design is described as off by default—admins choose whether it’s enabled and who can access it.
Pricing is comparatively straightforward: Claude Design is included at no extra cost within Pro. Max. Team. and Enterprise plans. with optional extra usage beyond subscription limits.. The model pricing for Opus 4.7 matches prior token rates, keeping the economics aligned with how Anthropic already sells access.
The strategic bet: owning more of the creative stack
Claude Design fits into a broader pattern.. Over the past months. Anthropic has moved from being mainly a model provider into offering more complete work tools: coding agents. knowledge-work assistance. office integrations. and agentic browser control.. Claude Design extends that trajectory into the visual layer—where product ideas are often turned into decisions.
The company’s financial momentum suggests it’s taking this expansion seriously, while the business risks are equally clear. Building an “application empire” while managing AI safety scrutiny, growing public resistance, and partner politics is a delicate balancing act.
The most likely near-term reality is that Claude Design won’t instantly replace incumbent tools for every team. But it may change who can prototype, how fast ideas become testable, and what “design time” even means for smaller orgs.
If that happens, the UI/UX category won’t just be about which tool produces the best pixels—it will be about which platform turns concepts into shippable product faster. And in that race, Anthropic is now trying to run from the first sentence all the way to production code.
Old iPad or Android tablet as a smart home panel: the setup guide
OpenAI exits: Kevin Weil and Bill Peebles leave as “side quests” end
Anthropic’s Mythos Preview: cybersecurity push aims to ease US tensions