Read-only design engineering for coding agents.
Memi gives Codex, Claude Code, Cursor, Grok Build, and MCP clients file-anchored interface evidence before they edit UI.
npx -y @memi-design/cli@2.6.3 diagnose . --json --no-write --fail-on noneNo account, API key, Figma file, global install, or daemon is required for the first audit.
| Repository | Role |
|---|---|
memi |
Core CLI, MCP server, GitHub Action, focused Agent Skills, and audit engine |
design-skills |
Governed catalog of 94 design, research, craft, generation, and Figma skills |
memi-studio |
Native macOS companion for supervised agent workflows |
design-sandbox |
Runnable Next.js proof environment for design audits and integrations |
audit-frontend-design— inspect interface risks before changing UI.remember-design-system— load compact product-system context.enforce-design-ci— gate pull requests with deterministic evidence.
Install the smallest workflow needed:
npx skills add memi-design/memi --skill audit-frontend-designmermaid-jam— local-only FigJam plugin for editable Mermaid and markdown diagrams.ripple-image-transitions— SwiftUI and Metal evaluation fork.chatbot— shadcn chatbot integration proof with Memi design CI.
Proof forks preserve upstream attribution and do not imply partnership.
The official CLI and Studio Homebrew tap lives at
homebrew-memi.
- Read-only inspection is the default.
- Findings include confidence, provenance, and file evidence.
- Deterministic checks rerun before a result is called verified.
- Existing design systems remain the source of truth.
- New workflow contributions land in skills before compatibility shims.
- Public integrations must be runnable, attributed, and maintained.
Documentation · npm · Discussions · Organization architecture