Thank you!
Your submission has been received!
Your submission has been received!
Oops! Something went wrong while submitting the form.
Tokens, module, column rules and measure written as structured context rather than a slide deck. The test is whether someone who isn't a designer can direct an AI to build a screen and have it come back on-brand. If the system only lives in Figma, they can't.
Your team generates code with AI and it comes out off-brand every time, so someone spends their afternoon correcting spacing and colour. The design system exists in Figma, which the agent writing the code cannot read.
A design system written only for people is now half a system. If AI tools are producing screens in your codebase, the rules need to exist somewhere those tools can read before they generate, not in a document someone checks afterwards.
That means the tokens, the module, the type scale and the actual constraints — what to do, what never to do — written as a plain file that lives with the code. Specific enough to be checkable. “Use spacing generously” is useless; “every vertical gap is a multiple of the module” can be verified.
This site runs on exactly that. There's a published rules file, it's the same file the build process works from, and it's linked publicly from the design system page.
Two to four weeks if a design system already exists. Longer if the underlying system needs defining first.
Does this replace our Figma library?
No. It's the same system expressed where the code gets written, so both halves stay in agreement.
Which AI tools does it work with?
Any that read files from the repository. It's a plain text file, not a tool-specific format.