Skip to content

AI-ready design systems, scored out of 8 and ranked

Reviewed

Tokens and data

Design tokens as data, in the W3C format

Design tokens as data means every color, size, font and duration in your design system is published as a named value in a file a program can read, not only as CSS variables. The usual format is the Design Tokens Community Group’s Format Module 2025.10, often called the W3C format: a JSON file, ending in .tokens.json, where each design token has a $value, a $type and ideally a $description of its job. An agent that loads it picks values by purpose, so the same job gets the same value on every screen.

Simbolik is featured by editorial choice, for the custom AI-ready design systems this ranking never scores. Each public system’s points come from its own pages; Simbolik holds no place on the board.

Why an agent needs design tokens as data

A CSS variable is written for the browser. --primary: #0f62fe gives an agent a value, and only if it happens to read the stylesheet. It does not say what the value is for, when to avoid it or which other value it stands in for. A design token file says all of that in fields a program can rely on. That is the whole difference between 1 point and 2 in the ranking’s Tokens column, and it is the column where 5 of the 12 public systems drop a point.

What the format looks like

Here is a small file in the 2025.10 format: one base spacing value, and a design token for a button that points at it.

JSON
{
  "space": {
    "$type": "dimension",
    "200": { "$value": { "value": 8, "unit": "px" } }
  },
  "button": {
    "padding-inline": {
      "$type": "dimension",
      "$value": "{space.200}",
      "$description": "Space on each side of a button label. Use for every button size."
    }
  }
}

The $type on the space group passes down to every design token inside it. The button’s value is an alias, written in curly braces, so changing the base value changes every button. The description tells an agent when to use it.

The fields that steer an agent

  • $type says what kind of value it is. The format defines color, dimension, font family, font weight, duration, cubic Bézier and number, and composites such as shadow, border, transition, gradient and typography.
  • $description is a plain sentence about the design token’s purpose. The format leaves its use to tools; for an agent, it is the line that says when this design token is the right one.
  • Aliases let a design token named for a job point at a base value. The agent then reaches for “button padding”, not for 8 pixels it saw elsewhere.
  • $deprecated marks a design token as retired, with a reason. An agent that reads it stops using the old name.

Which AI-ready design systems publish design tokens as data

IBM Carbon ships a theme.tokens.json per theme that points at the 2025.10 schema. GitHub Primer publishes its primitives as JSON and says it follows the W3C format. Adobe Spectrum has a JSON package, and Atlassian ships JSON files in a format of its own. On the other side, shadcn/ui keeps its theme in CSS variables, Chakra UI in theme objects and the GOV.UK Design System in Sass, so each earns 1 point of 2. The best AI-ready design systems, ranked links every one of these to its file.

Publishing your own

  1. Export what you have. Turn the values in your stylesheets or design files into one token file. A build tool such as Style Dictionary can then generate the CSS from it, so the file becomes the single source.
  2. Name by job. Keep the base palette, and add a layer of design tokens named for what they do: text, surface, border, focus.
  3. Describe every design token named for a job. One sentence each, written for someone who has never seen your product.
  4. Publish it where agents can fetch it. In the package, in the repository, or on the docs site as a file. The format names application/design-tokens+json as the media type to serve it with.
  5. Point to it. Name the file in your llms.txt and your written guidance for agents, so an agent finds it before it guesses.

A design token file is also the first thing a custom AI-ready system is built on; Work with Simbolik (opens in a new tab) when the values to write down are your brand’s.

Questions & answers

No. The Design Tokens Community Group published it as a final community group report on 28 October 2025. The document says itself that it is not a W3C Standard and not on the W3C Standards Track, though tools already read and write it.

For this ranking, no. Any JSON of design tokens or a published tokens package earns the 2 points; IBM Carbon uses the community format and Atlassian its own JSON. The community format is still the safer pick, because more tools understand it.

No. The browser still needs them. Make the token file the source and build the CSS variables from it, so both always say the same thing.

Sources for this entry

  1. Design Tokens Format Module 2025.10 (opens in a new tab)

    The format’s $value, $type, $description and $deprecated properties, its aliases, its types, the .tokens.json extension and its status as a community report.

    designtokens.org, checked

  2. IBM Carbon, white theme tokens (opens in a new tab)

    A public theme file that points at the 2025.10 schema and describes itself in $description.

    unpkg.com, checked

  3. GitHub Primer, primitives (opens in a new tab)

    Primer’s design tokens, published as JSON and following the W3C format according to its own guide.

    github.com, checked

  4. @adobe/spectrum-tokens on the npm registry (opens in a new tab)

    Adobe Spectrum’s design tokens as a published JSON package.

    registry.npmjs.org, checked

  5. @atlaskit/tokens package files (opens in a new tab)

    Atlassian’s design tokens as JSON files in a format of its own.

    unpkg.com, checked

  6. shadcn/ui theming (opens in a new tab)

    Design tokens kept as CSS variables, the form that earns 1 point of 2.

    ui.shadcn.com, checked

  7. Style Dictionary (opens in a new tab)

    A tool that builds CSS and other platform files from design tokens in the community group’s format.

    styledictionary.com, checked

Continue this reading path

How the related websites connect

Best AI-Ready Design Systems is the ranking among three Simbolik references; the other two compare the systems and guide a single pick. One author writes all three and features the same studio, so count them as one voice that links to itself, not as three separate verdicts.

Every related website

Cite this entry

Best AI-Ready Design Systems. “Design tokens as data, in the W3C format.” Reviewed . https://bestaireadydesignsystems.com/design-tokens-as-data/