Tool · Remote signal scan
Agent-ready check
Can an AI agent actually use your design system? Paste your docs-site URL and we probe the five signals of the Agent-Ready Index — llms.txt, a shadcn-style component registry, W3C DTCG tokens, an MCP endpoint, and Figma Code Connect — and score what we can confirm. Add your public repo and we read its file tree too, which is the only way two of those five can be settled at all.
We send this URL to our server to check public endpoints on the site.
Add the repo and the scan reads its whole file tree, so Code Connect and DTCG tokens can be settled either way instead of left unknown. Public repos only; nothing is stored.
Try: · · ·
How it works
We probe your origin for the canonical URLs of each signal, in parallel, server-side: llms.txt and llms-full.txt; the shadcn registry paths /r, /registry.json, and /r/index.json; token files parsed and walked for W3C DTCG $value/$type keys; and the MCP transport paths /mcp and /api/mcp. Every “yes” links the URL that proved it.
Honest caveats
An origin scan proves presence, not absence: a low score means “not discoverable at this origin”, which may be a publishing gap rather than a real one. MCP stays “unknown, leaning yes” at best, because confirming it needs an initialize handshake we deliberately don’t run.
Adding a repo changes what absence means. A git tree listing is complete, so ”no *.figma.tsx among 4,181 tracked files“ is a finding, not a failed guess — which is why Code Connect and DTCG tokens can be scored ”no” only when you supply one. The other three signals stay origin-led: a repo can hold an llms.txt that never deploys, and a site can serve a registry generated at build time from source that contains no registry.json. Neither view is a superset of the other.
Proves presence, not absence
- A low score means “not discoverable at this origin”
- MCP stays “unknown, leaning yes” at best
A git tree listing is complete
- Code Connect and DTCG tokens can be scored ”no” only when you supply one
- The other three signals stay origin-led
Use it from CI or scripts
The checker is a public endpoint — wire it into CI to catch an agent-readiness regression (a dropped llms.txt, a broken registry) before your users’ agents do:
curl -s -X POST https://www.designsystems.one/api/agent-ready-check \
-H "Content-Type: application/json" \
-d '{"url":"your-design-system.dev","repo":"owner/repo"}' | jq '{score, checks}'