A working three-panel tool: system health, a token and component explorer, and an Ask Aero interface wired to the Anthropic API. Most designers talk about AI and systems. This is what it looks like when you build it.
The mini Aero dashboard is a proof of concept and a working demo — both at once. Its job is to show what a design system looks like when it's been instrumented for AI: queryable, observable, and able to answer questions in natural language.
It's also the prototype for Stage 5 of the agentic design system roadmap. Building it first makes the vision concrete and testable before any engineering team commits to the full MCP infrastructure.
[ Fill in: What specifically prompted you to build this. What you wanted to prove. What you learned from building it that changed the roadmap. ]
Component coverage, token adoption, documentation completeness. A live view of where Aero is strong and where it's drifting.
[ Fill in: Specific metrics shown, data sources, refresh logic. ]
Browse, filter, and inspect Aero tokens and components. Values, usage rules, decision rationale — all in one place.
[ Fill in: How data is pulled from Figma MCP. What filtering/search looks like. ]
Natural language queries against the design system. "Which token should I use for a destructive action?" gets an answer grounded in Aero's actual decisions.
[ Fill in: How the Anthropic API is wired. What context is passed. Example queries and answers. ]
All development done via Claude Code in the CLI. [ Fill in: What that workflow looked like in practice. What surprised you. What you'd do differently. ]
Figma's Model Context Protocol integration allows the dashboard to query design tokens and component data programmatically. [ Fill in: What the MCP connection enables. What data flows from Figma and in what format. ]
The Ask Aero panel is powered by Claude, with the Aero token documentation passed as context. [ Fill in: The prompting strategy. What context is injected. How you handle hallucinations or out-of-scope questions. ]
[ Fill in: Deployment details. How updates are pushed. Any CI/CD or environment variable setup worth noting. ]
[ Fill in: What the dashboard informs about the full MCP roadmap. What you need to build next. How this becomes the interface layer for Stage 5 of the agentic design system plan. ]
[ Fill in: What you'd add to the dashboard if you had more time. What you learned from building it about what designers actually need from AI-instrumented systems. ]
Let's talk