Independent audit · 2026-09-04
Every system in the gallery scored against five concrete signals that determine whether Cursor, Claude Code, or v0 can do useful work with it. First-party evidence only — community wrappers, third-party MCP shims, and generic "supports Figma" claims don’t count. Re-audited quarterly.
The systems closest to agent-ready as of 2026-09-04. Ties broken alphabetically.
An independent index shouldn’t grade its author, so designsystems.one isn’t a row above. Scored here on 2026-09-02 against the same five rules — a first-party artifact on the canonical domain, or it doesn’t count: 4/5.
First-party MCP server over Streamable HTTP (JSON-RPC 2.0), no auth, also at /mcp.
/api/mcpllms.txt and llms-full.txt on the canonical domain.
/llms.txtToken generator emits W3C DTCG format ($value / $type / $description).
/api/tokens?format=w3cshadcn-spec registry index at /r, items installable via `npx shadcn add`.
/rNo Figma Code Connect mappings are published for the registry components.
Two views of the same audit: which signals the 37 systems actually ship, and how far the field is from fully agent-ready.
llms.txt is the cheap race; the registry lane is still nearly empty.
One system reaches 4/5 (Carbon); none reaches 5/5 — fully agent-ready doesn’t exist yet.
Each system gets a score from 0 to 5: one point per signal where the maintainers themselves publish the artifact at a first-party URL.
/llms.txt on the canonical docs domain returns content.$value / $type).npx shadcn add <url>..figma.tsx) shipped publicly."Unknown" isn’t a guess — it’s the honest answer when we couldn’t verify a first-party artifact at the canonical URL within the audit window. It does not count as a "no". Code Connect in particular is under-disclosed and may be the most-undercounted signal here.
If we got something wrong about your system, the gallery breakdown for each system links the source — open an issue and we’ll re-audit.
Sorted by score, ties broken alphabetically. Top score in the audit is 4/5.
Unknown is not absence. The confirmed score measures published artifacts; source coverage shows how much can be independently checked. Inspect evidence and dates.
$value / $type conformance explicitly. The spec only stabilized in October 2025, so many systems are mid-migration.The dataset is free to reuse under CC BY 4.0 — attribute DesignSystems.one and link back to the index URL. JSON and CSV downloads above are stable per audit date.
DesignSystems.one. (2026). Agent-Ready Design Systems Index. https://www.designsystems.one/ai-ready/systems · audited 2026-09-04
Run the same five checks against your own docs in about ten seconds — free, no signup. For the full picture (Code Connect, token architecture, and a sequenced remediation roadmap), the Agent-Readiness Audit applies this exact methodology to your system: $3,500, two weeks, fixed.