Sibling contract
AI Readiness
AI readiness is the 6th maturity axis (zeroheight 2026). A design system is AI-ready when it exposes machine-readable context that AI coding agents can consume: tokens, rules, discovery endpoints, and verification. This contract defines the checks that score it.
Version 0.1.1 · provisional · machine export
Source authority
- Primary
- zeroheight 2026 Design System Maturity Model: AI Readiness axis
- Definition
- Whether your design system is the default context that AI tools build from, or whether AI is silently working around it
- How-to
I made my design system LLM-readable. Here's exactly how.
- DLS Lead
The six maturity axes
01 Foundations
02 Documentation
03 Governance
04 Adoption
05 Measurement
06 AI Readiness (new)
AI-readiness indicators
- Machine-readable token file (DTCG/JSON) exposed at a discoverable URL
- llms.txt brief present at /.well-known or /llms.txt
- agent.json discovery document at /.well-known/agent.json
- MCP endpoint exposing tools/list for AI agents
- DESIGN.md present (the DESIGN.md format)
- Token documentation: tokens carry $description for agent context
- Component contract: machine-readable component schema
- Sitemap.xml present for crawlability
- robots.txt permissive for agent crawlers
- Open Graph + Twitter meta for social/share context
Verification
10 checks. PASS=1, WARN=0.5, FAIL=0. Score = (points/10) × 100. A≥90, B≥80, C≥70, D≥60, F<60.
01 r01 · Machine-readable token file exposed (DTCG/JSON) Pass Token file found at a discoverable URL Warn Token file found but not in DTCG format Fail No machine-readable token file detected
02 r02 · llms.txt present Pass /llms.txt returns agent-facing brief Fail No /llms.txt found (agents have no orientation brief)
03 r03 · agent.json present Pass /.well-known/agent.json returns discovery document Fail No /.well-known/agent.json (agents have no discovery endpoint)
04 r04 · MCP endpoint responds to tools/list Pass MCP endpoint found and responds with tool list Warn Endpoint exists but did not respond to tools/list Fail No MCP endpoint detected (agents cannot call tools)
05 r05 · DESIGN.md present (the DESIGN.md format) Pass /DESIGN.md found; follows the DESIGN.md format Fail No /DESIGN.md (missing the machine-readable design brief)
06 r06 · Token documentation: tokens carry $description Pass Tokens found with $description metadata Warn Some tokens have $description, others do not Fail Tokens found but no $description (agents lack context)
07 r07 · Component contract (machine-readable schema) Pass Machine-readable component schema detected Warn Component documentation found but not machine-readable Fail No machine-readable component contract
08 r08 · Sitemap.xml present Pass /sitemap.xml found (site is crawlable) Fail No /sitemap.xml (agents cannot discover pages)
09 r09 · robots.txt permissive for agents Pass /robots.txt found and allows agent crawling Warn robots.txt found but restricts some agents Fail No /robots.txt (crawling rules undefined)
10 r10 · Open Graph + Twitter card meta tags Pass OG and Twitter card meta tags present Warn Partial social meta (only one of OG and Twitter) Fail No social meta tags (missing share context)
Validation
Designesy readiness engine: HTTP probing of machine-readable artifacts. Method: HEAD/GET requests against the target origin for each known artifact path. Browser-only checks: None: all 10 checks are HTTP probes, no browser needed.
Open questions
- r01 (token file) probes common paths (/tokens.json, /design-tokens.json): custom paths may be missed
- r07 (component contract) is heuristic: no universal standard for machine-readable component schemas exists yet
- r04 (MCP endpoint) probes /api/mcp: custom MCP paths may be missed
- The score does not measure token quality: only presence. A badly structured token file still scores PASS on r01
Run it
Score any URL for AI readiness:
The AI Readiness contract reads AI readiness, the sixth axis of zeroheight's maturity model, with an automated scan of the site in place of a questionnaire. Machine export.