The design system is becoming agent context
Coinbase reports that connecting production components to Figma helped coding agents follow its design system and reduced token costs by an average of 22.5%.
average reduction in token costs reported by Coinbase when using Code Connect with coding agents
The Story
Coinbase’s case study is a useful counterpoint to vague claims about AI productivity. Its focus is not on generating more screens, but on giving agents better access to the components and rules that already exist.
That reframes a design system as operational knowledge: naming, component relationships, implementation guidance, and intent. The stronger the context, the less an agent has to guess—and the less expensive those guesses become.
For Studio Wensday, the lesson is concrete: when building a client-facing system, document the “why” alongside the “what.” A component library that explains its decisions will travel further into code, content, and future AI workflows. ([figma.com](https://www.figma.com/blog/ai/?utm_source=openai))
Studio Wensday Angle
A design system is not only a visual library; it is a context layer for people and machines.