The AI-Native Design Stack: What Elux Space Uses in 2026

People ask what tools we use. The tools are not the interesting part. The interesting part is how they connect and what that connection enables. This is the full stack breakdown, including cost.

The full toolchain

Claude Code is the prototyping and implementation engine. It builds working interfaces from briefs. It connects to Figma via MCP to push interfaces to the canvas and pull design context into code generation. It is used for every client sprint at Elux Space from day one. Cost: Claude Max subscription, approximately $100 per month per seat.

Figma with MCP is the design system and review environment. Interfaces built in Claude Code land here as editable frames. The design system documentation, component library, and token map live here. Clients review here. Changes annotated here feed back into Claude Code via MCP. Cost: Figma Dev or Full seat, $25 to $45 per month per seat depending on plan.

Webflow handles marketing sites, client blogs, and content-driven web products. Design tokens from Figma map to Webflow variables. The marketing site and the product UI share a visual language without manual maintenance. Cost: $23 to $39 per month per site on standard plans.

paper.design is used for specific client projects where HTML/CSS output is the primary deliverable and the design-to-code gap needs to be zero. Not every engagement uses it. When it fits, it fits better than anything else in the stack for that specific job.

How the tools connect

The pipeline flows in one primary direction with a feedback loop. Brief enters Claude Code. Working interface exits Claude Code and enters Figma via MCP. Client reviews Figma. Feedback re-enters Claude Code via the designer's prompts informed by MCP context. Refined interface exits Claude Code and updates Figma. Final design system in Figma informs Webflow variables. Marketing site goes live in Webflow.

No handoff documents. No manual token reconciliation. No spec sheets that become outdated the moment a decision changes. The pipeline carries context through each step automatically because the tools are connected rather than sequential.

What this costs vs a traditional agency setup

The AI-native stack for a three-person design agency costs approximately $500 to $700 per month in tool subscriptions. A comparable traditional agency setup, including project management tools, design software, prototyping tools, and developer handoff platforms, costs roughly the same amount but requires three to five times as many hours to produce equivalent output because the tools are not connected.

The real cost difference is in hours billed to clients. Traditional agency engagements bill more hours because more manual work exists between phases. AI-native engagements bill fewer hours because the connections handle the manual work. For clients, this means a shorter engagement for a comparable deliverable. For the agency, it means more projects can run simultaneously without proportional headcount growth.

What this enables that wasn't possible 18 months ago

Eighteen months ago, pushing a working interface into Figma required a developer to produce screenshots and a designer to manually recreate the layout as Figma frames. That step took half a day at minimum. Today it takes minutes via MCP. That one change, a single step in the workflow, is worth approximately two days per sprint.

Multiplied across a year of client work, those two days per sprint represent a significant portion of the capacity gap between a three-person AI-native team and a ten-person traditional team. The capacity is not in headcount. It is in the connections between tools.