AI project - 03
Started with one client's component backlog. Headed towards an end-to-end orchestrator.
Status
Deployed for client
stack
Figma MCP, Variables, Storybook
type
Client-scoped pilot
the problem
Design systems fall behind product work constantly — a team needs a new component variant now, but touching the shared library (McKesson Mobile Companion Design System) means governance and delay. The usual workaround is a designer building components and variants locally, disconnected from the design system, creating the drift the moment it ships.
Today - Client pilot
Currently running for McKesson's Mobile Companion Design System: a personal pipeline that builds new components in a working file, composed from the shared library's existing primitives and bound to its tokens and variables — never touching or modifying the source library itself. Every request is checked against what's already built before anything new is created, and each component ships with its own documentation drafted directly into Figma.
Where its headed - planned
The next version generalizes this into an end-to-end, product-agnostic design system pipeline — one orchestrator coordinating specialized sub-agents, each owning a stage of the system's lifecycle:
Component creation
New elements built directly in Figma
Tokens and Variables
Generated, not hand-authored
Foundations
Color, type, spacing, elevation systems generated
Figma → Storybook
Publishing straight to the dev library
Documentation
Generated per component, automatically
Accessibility prep
Every element re-validated against standards
Roadmap is built and progressed upon — shown to describe direction and the end goal.
relevant tags
Design systems
Figma mcp
multi-agent orchestration
Ds-ops
design-dev handoff