Questions about the AI-ready design systems ranking

Four AI-ready design systems share first place with 8 of 8. IBM Carbon, Atlassian, Adobe Spectrum 2 and Ant Design each earn full points on design tokens, agent access, docs and guidance. Pick between them by your stack and the kind of product you build.

These three AI-ready design systems, shadcn/ui, Chakra UI and Shopify Polaris, each score 7. A tie shares its place, so the next system, GitHub Primer, is eighth.

No. Simbolik makes it, and each page says so beside the Site’s name and in the footer. The About page names Simon Martin, who runs Simbolik. The points for each AI-ready design system are kept to its own pages so you can check each one yourself.

Each public system gets 4 ratings: design tokens, agent access, docs and guidance. A Yes earns 2 points, a Partly 1 and a No 0, so 8 is the most any system can score. Systems on the same score share a place, and the next place skips ahead by the number tied. The page on how AI-ready design systems are scored gives the rule for each column.

Rate your AI-ready design system on the same 4 columns, add up the points, then fix the cheapest gaps first. An llms.txt and a page of rules for agents are text files, so Docs and Guidance are usually the fastest 4 points. Design tokens as data take longer, and an MCP server is the biggest job, so it comes last. The page on raising your score walks through each column.

Nobody can price an AI-ready design system honestly without seeing the system, because the job depends on what already exists. The cost guide sorts the work from cheapest to dearest, starting with written guidance and docs as text. Simbolik publishes no price; it states one in a one-page proposal agreed before work starts.

That depends on the gaps in the AI-ready design system. A team working in-house should score its system and run a pilot first, then plan from the list of gaps those produce. The cost guide has the order in which the parts usually go, and the pilot guide shows how to run one with a fixed end.

Then it comes first, and it is the largest part of the job. Everything else an AI-ready design system needs is built on it: design tokens, typed components, docs, guidance and checks. The ranking can only score a system that exists, so the cost guide lists this step on its own.

No AI-ready design system on the board is documented as tied to one agent. The official servers use MCP, an open protocol many agents speak. shadcn/ui, for one, documents its server for Claude Code, Cursor, VS Code and Codex. Check that your agent is on a server’s list before you plan around it.

Usually, as long as every copy keeps the license notice. Most of the ranked AI-ready design systems license their code for reuse, though a few have gaps: Shopify states no license for the Polaris web components on the pages checked, and Salesforce’s tokens package points to a license file of its own. The licenses do not hand over the maker’s brand: IBM Plex reserves its font name, for one. Fonts and icons have terms of their own. The licenses guide for AI agents reads them one part at a time. It is not legal advice.

Open the entry of the AI-ready design system in the directory. It links the public page each rating rests on, such as the system’s own docs, its official GitHub or its npm page, and shows the day that page was opened. A third-party list is never used. A part that cannot be found on those pages is not scored as a Yes.