Agentic Design Systems: Foundations
A 7-week live course for designers who want their system to ship real code. You'll build tokens, components, and an agent crew that builds, tests, and deploys, while you hold every gate. Zero code experience needed on day 1.
We are a Figma trusted partner. Join, give us the Gmail on your Figma account, and we help you apply for the free plan.
Pick the intake that fits your calendar.
Same course, same price ladder. Each cohort has its own dates, time, and seats. Full ones take a waitlist, not a payment.
AI has no memory of your decisions.
Every prompt starts from zero. The corner radius you fought for, the contrast pair that passed the audit, the spacing scale your product breathes in: the AI knows none of it. So it guesses. It averages the internet and hands you everyone's product. Which is no one's.
Ask it to design your screen in Figma and it comes back looking finished. Open the layers and there is nothing underneath. No components, no tokens, no rules. Nothing is connected to anything, so nothing can change. Thirty seconds to generate. Weeks to ship. Shipped wrong.
There is a worse version, and it's the popular one: skip design entirely and let AI go straight to code. Now your product's decisions live in files only the AI has read. You want to change a button, you have to ask. You want to know why the blue changed, you have to ask. It is your product, and you are a visitor in it.
A design system turns the file into shared memory. Your decisions live in Figma as tokens, components, and rules, written where both of you can read them. And the loop runs both ways. You retune a token in Figma, and the agent carries it into shipped code. The agent builds a screen in code, and it lands back on your canvas as instances of your components, where you can inspect it, correct it, approve it.
Figma stays the source of truth. Not because it is sacred, but because it is the one place where everything is visible, reachable, and yours. The AI does the labor. You keep the file.
That gap is what this course closes.
Week by week
The ground
- Day 1 · What a system is. You inherit a real, messy product file and hunt its inconsistencies. Rules that repeat, and why 14 button styles happened to a talented team.
- Day 2 · Meet the third reader. Designers read your file, developers read your file, and now AI reads your file. How it reads, what it needs, and your first look at AI-written code.
The whole machine
show the sessions
- Day 1 · The full design system ecosystem, from UI kit to component library to the real thing, plus the layer stack that connects Figma to shipped code. You connect your first tools live in class.
- Day 2 · Onboard your dream team. The pitch that wins adoption: value statement, reading the room, honest proof.
Inventory and tokens
show the sessions
- Day 1 · Interface inventory, design system roadmap, and how to run the build as a project.
- Day 2 · Design tokens from zero: tiers, aliasing, naming, light and dark. Then you build your system's inventory base in Airtable. Keep that base. It comes back later, and it grows into something bigger than a catalog.
Tokens shipped, components built
show the sessions
- Day 1 · Token implementation, the full pipeline: Figma to JSON to Style Dictionary to Git to Storybook. You watch a color change travel from a Figma variable into live CSS.
- Day 2 · Components, the brick before the wall. The 5 property types, composable subcomponents, and the AI-readable checklist: naming, tokens, states, documentation. Your components carry their own behavior, so you never teach AI the same interaction twice.
From node to crew
show the sessions
- Day 1 · Node to code. Claude Code reads your Figma component through MCP and builds it, driven by a skill you configure.
- Day 2 · One agent to a crew. The agent loop, workflows vs agents, orchestrator and workers, and the human gate that keeps you in charge.
Mission control and release
show the sessions
- Day 1 · Run your system from a board. Your Airtable base becomes a control plane: an 11-state component lifecycle, a 4-agent crew (manager, engineer, tester, devops), and 3 gates only you can open.
- Day 2 · Releasing a design system. Test it 5 ways, audit it with AI through 4 lenses, document it, version it, and publish a real docs site.
Apply the system everywhere
show the sessions
- Day 1 · Design like a pro on your own system. The 4 mistakes that quietly break systems, flow design with prototype linking, frame naming AI can parse, and setting up a real project repo with agents and preflight checks.
- Day 2 · The crew ships the flow. A flow-identifier agent maps your screens, a parity checker guards your system's boundaries, the crew builds, tests, and you preview, approve, and merge. Then you grade your agents. We close with your capstone brief.
Ship a system, ship a flow.
After the course, you build a small design system from zero on a fresh fictional brief (you pick 1 of 3), then push one full user flow through it: designed, agent-built, tested, gated, merged.
Submit it in Google Classroom and demo it live in a scheduled review: open the board, open the repo, click through the flow. We review every submission by hand, and you earn the TPS certificate for Agentic Design Systems: Foundations on a pass.
has completed the Agentic Design Systems: Foundations
By the end of 8 weeks, you can:
Audit any messy product file and turn it into an inventory, a roadmap, and a plan leadership will fund.
Build a 3-tier token architecture with theming modes, wired from Figma all the way into shipped CSS.
Design components that AI reuses instead of reinvents: named, tokenized, documented, with their interactions built in.
Run your system from a live dashboard where AI agents build, test, and deploy components, and nothing moves past your 3 human gates without you.
Ship a full user flow from Figma to working, animated code through your own agent crew, then merge it yourself.
Grade your agents monthly and decide who gets promoted. Yes, really.
And underneath all of it, one skill that outlasts every tool: you'll know how to stay the disciplined input, because AI scales whatever you feed it. Garbage in, garbage scaled.
You learn by doing the job, in class.
What comes with your seat
Both answers save you money.
You're a UX or product designer with solid design judgment. You know a good interface when you build one.
You've watched AI tools generate messy, inconsistent UI and thought “there has to be a way to control this.”
You want to be the person in your org who runs the design system, and the agents that maintain it.
You've never written production code. That's fine. That's the plan.
You want a Figma basics class. I'll assume you already use Figma better than most of your leadership.
You want to become a software engineer. You'll read code and direct agents that write it, but engineering is a different course entirely.
You can't attend live. The activities run on real files with real time pressure, and recordings alone won't get you there.
What you need before day one
One discount at a time. If two could apply, you get the lower price. They don't stack.
Asked before you asked.
I can't code at all. Really, at all. Will I survive?
Do I need a powerful computer?
What if I miss a session?
How much time outside class?
Is the certificate automatic?
Foundations is open. Four steps to a seat.
Apply and tell us how you work with AI today.
Sign the consent for the cohort.
Pay, then upload your payment proof.
We verify by hand and confirm your seat.