The agentic design system platform. Your design and engineering knowledge, connected into intelligence every agent can use β plus agents that maintain it.
Joined April 2016
- Tweets2.7K
- Following778
- Followers3.1K
- Likes8.8K
AI is changing how we build products. So what happens to design systems?
We asked the people figuring it out at WaysConf.
Hereβs what they told us π
Design system maintenance eating your week?
Let AI agents handle it.
Live webinar, Oct 21:
β Connect from Cursor, Claude, Codex & more
β Scoped MCPs for every team
β Agents that edit tokens, docs & more
Register π bit.ly/supernova-mcp
Canβt make it? Get the recording.
Your AI can write code. But does it understand and maintain your design system? π
Join us Wed, Oct 21 for a live look at Supernova MCP:
β Cursor, Claude, VS Code, Codex & ChatGPT
β Scoped context for every team
β Automated design system maintenance
β New Write MCP tools
Live demos + real use cases + Q&A.
Register here: bit.ly/supernova-mcp-webinar
We made learning Supernova shorter than your standup.
Supernova Academy π Short videos, real workflows:
β Keep docs updated with agents
β Turn token changes into PRs
β Connect your DS to Cursor & Claude Code via MCP
β Measure adoption
What should we cover next?
supernova.io/academy
Your AI agents work with your design system. Now you can see how. π
MCP Analytics is live in Supernova π
See:
β What agents change + how often
β Which tools they use most
β Whoβs connecting
β Which contexts drive usage
β How usage changes over time
What would you check first?
Whatβs the strangest thing an AI agent has invented in your codebase?
β A new spacing scale?
β A button variant nobody designed?
β A sixth shade of blue?
AI agents don't ignore your design system. They invent their own.
And the worst part? It can still pass code review.
The fix: give agents access to your real design system.
Supernova's Editor MCP connects your design system to Claude Code, Cursor, Codex, and VS Code with Copilot, so agents can read and update your actual tokens, components, and rules.
Try it β‘οΈ supernova.io/blog/connect-deβ¦
Most design system MCP servers can read. Very few can write.
Deprecate every Button variant? Clean up 10 docs pages?
Without write access: a day of clicking. With it: one prompt.
We compared 5, ours included, on what they can actually do β‘οΈ supernova.io/blog/design-sysβ¦
Do agents kill design systems, or prove their point?
That was the question we took to WaysConf in KrakΓ³w last week.
One round table. Sharp debate. A room full of design system people.
The verdict? Agents donβt replace design systems. They make them matter more.
πͺ Thanks to everyone who joined us!
Supernova is at @WaysConf today! π« in KrakΓ³w, Poland
Weβre bringing our agentic design system platform. With Editor MCP, AI agents can now build, edit & maintain your design system.
π Booth next to LeΕΌaczki Lounge
π€ Roundtable: Do agents kill design systems, or prove their point? | 14:55β15:50
Come say hi, grab some swag, and challenge us with your hardest questions!
Design system teams are understaffed. Expectations aren't slowing down. Usually, 2-5 people hold the system for hundreds.
So we built Editor MCP β real write access to your design system from Cursor, Claude, VS Code, Codex, or ChatGPT β to let agents document and manage your design system.
Automate up to 90% of your daily maintenance.
Live in Public Beta, free. π
Brand inconsistency rarely happens all at once.
It's quiet. One team uses hex values from memory. Another exports from a two-year-old Figma file. A localization decision becomes a permanent fork. By the time leadership notices, the fixes are expensive.
The teams that stay consistent at scale aren't more disciplined; they just made consistency the default output of their tools.
That's what a well-built design system does for a brand: it moves the brand from something teams have to remember to something the system enforces automatically.
Our latest post covers how to build that foundation: token layers, controlled flexibility, and keeping design and code in sync.
Read it here: supernova.io/blog/building-aβ¦
We just shipped AI context management π₯π₯π₯
Your AI agents are only as good as the context you give them. The common approach so far? Hand the agent your entire design system via a single MCP and let it sort things out. It somewhat works, but the output drifts, quality suffers, and you pay for the excess in AI tokens.
We built a better way.
Now you can define exactly what each team's agents need to know (design tokens, component APIs, rules, docs) and publish it as a focused, scoped MCP endpoint they plug into directly.
β
Scope context per team, workflow, or agent type
β
MCP now includes semantic search and full code component knowledge
β
Distribute as a ready-to-use plugin for Cursor, Claude Code, or Codex in seconds
β
Update once and every connected agent gets it instantly
Learn more: learn.supernova.io/latest/reβ¦
A design system that can't speak business gets cut.
The ones that survive don't just ship components. They tie every roadmap priority to an outcome that leadership already measures.
Faster launches β component coverage for your most-used patterns.
Cost reduction β eliminating duplicated UI across repos.
Market expansion β theming architecture that makes new regions a config, not a rebuild.
We broke down how to make this work in practice.
Read our blog π
supernova.io/blog/align-desiβ¦
Supernova.io π« retweeted
π Martin Ε karbala, Staff QA Engineer @supernova_io
π Thursday, May 28 at 12:40 pm
π webexpo.net/prague2026/sessiβ¦
#TestAutomation #AIInTesting #QualityEngineering
Your sign to go to @webexpo Β ππ
Supernova is giving the community 1,000 CZK OFF tickets.
Use code: thanksSupernova β‘οΈ webexpo.net/
Come say hi at our booth! Weβd love to meet you!
