Make your design system AI-ready.
Stop your AI from inventing your design system.
Lyse scores it 0–100, every commit.
Leading design systems, audited — Lyse scores the design systems shipped by these teams.
Live scores
Even the best design systems drift.
200+ design system audits run every day on Lyse, across six axes — including AI-readiness. Open formula, scored nightly.
How it works
Three surfaces. One install.
CLI for your terminal. GitHub Action for every PR. MCP server for your AI agents.
Run the audit
Lyse reads your tokens, components, and stories. No config. No API key. Health Score in 30 seconds.
Score every PR
Lyse posts a Health Score on every pull request. SARIF lands in your Security tab. Drift can't ship silently.
Brief your agents
Lyse runs as an MCP server. Cursor, Claude Code, and Codex query it before writing code. No hallucinations.
Features
Everything your DS team needs. Nothing you don't.
Lyse handles the audit. Your team keeps every design decision.
Hardcoded color in Button.tsx
background: "#3B82F6"
3 occurrences across this component
Safe auto-fix
Lyse rewrites hardcoded values into your design tokens. Six safety guards before anything ships.
Health Score · 74/100
1,247 findings tracked across every commit
tokens/no-hardcoded-color · auto-fixed · 2h ago
components/shadow-native · resolved · Yesterday
a11y/essentials · 2 fixes applied · 3d ago
The Health Score
A single number, 0–100. Six axes — including AI-readiness. Open formula, calibrated against 70 repos.
Next: Autonomous
6 points · 80/100
Reached Quantitative
Last week · 74/100
Reached Defined
March 12 · 56/100
Foundational → Autonomous
Every score maps to a CMMI-style tier. Climb from Foundational to Autonomous, one commit at a time.
Built for AI agents
Your AI agents stop guessing your design system. Lyse generates AGENTS.md and runs as an MCP server.
Trust
Lyse audits. You decide. Always.
No black box. No silent rewrites. Every drift, every score, every fix — auditable, reversible, yours.
Local-first
Lyse runs entirely on your machine. Your source code never leaves your laptop. No telemetry, no API key, no account.
Deterministic outputs
Lyse uses an open, versioned formula. Same repo, same score, every run. A determinism gate runs in CI.
Human-in-the-loop
Lyse fix only proposes codemods — you review the diff like any PR. Six safety guards before anything ships.

Now machines read your design system too.
Your design system is now read by machines. We're building the open infrastructure that makes it AI-ready.
FAQ
Everything you need to know.
What DS Leads ask us in discovery calls.
+What does the Health Score actually measure?
Six weighted axes — tokens, components, accessibility, stories, naming, and AI-surface readiness. Each axis aggregates specific rules with transparent formulas. Plus a CMMI-style maturity tier. Open formula in docs/health-score.md.
+Will Lyse work with my tokens?
Yes. Lyse reads Style Dictionary, Tokens Studio, Tailwind v4 @theme, DTCG JSON, and CSS variables. If your build can find them, Lyse can too.
+Will my repo's score show up on the public Score Live?
Never. The Bench audits open-source repos only. Your internal score stays on your laptop and in your CI. Lyse Labs never receives it.
+How do my AI agents actually use Lyse?
Two paths. Static: `lyse agents > AGENTS.md` writes a contract your IDE reads. Live: `lyse mcp` runs an MCP server that Cursor, Claude Code, and Codex query — audit_file + suggest_fix in real time.
+My design system isn't perfect yet. Will Lyse just shame me?
No. Lyse renormalizes when an axis doesn't apply — no Storybook? That axis is N/A. You get a partial score plus a CMMI tier (Foundational → Autonomous) to track progress over time.
+Is Lyse going to be around in 2 years?
The code is AGPLv3 — fork-proof. Even if Lyse Labs disappears, you keep the tool. No VC raise, no exit pressure. Commercial license funds development. Building in public.
Make your design system AI-ready. Score it in 30 seconds.
Stop your AI from inventing your design system.