A design system built for AI to use, and for people to trust.
Ten layers of constraint, all enforced in code: accessibility, ethics, brand, motion, agent rules. Nothing merges until it clears all ten, so shipping at AI speed never costs you compliance, brand, or trust.
What ships today
Where to start
Four entry points into the Docs tab. Pick the one that matches what you're here for.
Copy this prompt to create your own fork of BADS
Paste it into Claude Code exactly as written. It interviews you for the brand, the audience, and the tone, then walks the whole onboarding — so nothing needs filling in first.
I want to create my own design system by forking BADS (https://github.com/matthewlarn/behavioral-agentic-design-system). Interview me one question at a time — what the product or client is called, what it does and who uses it, the brand colors and typefaces if they exist yet, and the tone the interface should take. Do not assume any of it and do not start editing files until I have answered. Then walk the onboarding in CLIENT_ONBOARDING.md end to end: run the rename with the display name and slug I give you, point the fork at the published @behavioral-agentic-design-system/core package rather than a vendored copy so it keeps receiving updates, replace the placeholder tokens with my real brand values and regenerate the CSS through the token build, fill in the brand voice and BRIEF files, and seed the exemplars and real-content pool. When you are done, run the accessibility and drift checks and show me what is still a placeholder.