memi

memi

Design QA skills for coding agents.
Audit real interfaces, remember their design systems, and stop UI regressions before merge.

Package: @memi-design/cli · Homepage: memoire.cv · MCP Registry: io.github.sarveshsea/memi

memi turns a real product codebase into evidence agents can trust before they edit UI: tokens, components, routes, screenshots, UX risks, interface craft scores, shadcn registries, research specs, and deterministic CI gates. Works with Grok Build (Grok 4.5), Codex, Claude Code, Cursor, Hermes, OpenCode, OpenClaw, and any MCP client.

Compatibility: shadcn registry and v0 design systems.

---

Try it on a real interface

npx -y @memi-design/[email protected] diagnose . --json --no-write --fail-on none

That runs a read-only, deterministic audit against the current repository and returns file-anchored evidence. No account, API key, Figma file, global install, or daemon is required.

To make the same workflow available whenever an agent edits UI:

npx skills add sarveshsea/memi --skill audit-frontend-design

Then ask: "Audit this frontend before editing it. Prioritize the five fixes that will matter most to users."

The skill runs the pinned CLI only when repository evidence is needed.

Four focused skills

Job	Install
Find accessibility, token, hierarchy, state, and responsive issues	npx skills add sarveshsea/memi --skill audit-frontend-design
Load compact design-system context before UI work	npx skills add sarveshsea/memi --skill remember-design-system
Add deterministic design checks to pull requests	npx skills add sarveshsea/memi --skill enforce-design-ci
Build and verify native SwiftUI interfaces	npx skills add sarveshsea/memi --skill build-swiftui-interface

Browse the skills on skills.sh or use the full memoire-design-tooling router for Figma, MCP, research, scaffolding, and registry workflows.

Focused mirror repositories remain available for directory discovery, but sarveshsea/memi is the source-of-record install.

Every CI finding cites file:line and re-runs identically. No LLM is used in the enforcement path.

Machine-readable index: llms.txt

---

For humans — five-minute proof

npm i -g @memi-design/cli

memi agent brief . --intent "Improve this interface" --detail compact --json
memi diagnose
memi ux audit --json
memi craft audit --json
memi tokens --from ./src --report
memi scaffold component EvidenceCard --level organism --json
memi ios brief --intent "Build an accessible settings screen" --detail compact --json
memi ios scaffold Settings --kind screen --module AppModule --json
memi shadcn export --out public/r
memi agent install universal --project .
memi mcp start --no-figma

That loop gives you compact web and Apple-platform briefs, app-quality findings, UX risks, craft critique, extracted tokens, dry-run web and SwiftUI file scaffolds, shadcn registry output, installable Agent Skills, and a Figma-independent MCP server.

Public proof repo: sarveshsea/design-sandbox — Next.js 16 + Tailwind 4 + shadcn wired with MCP, skills, and memoire.agent.yaml.

---

Grok Build (Grok 4.5) — recommended setup

Grok Build is xAI's agentic coding CLI powered by Grok 4.5. memi is the design-memory layer that stops Grok from guessing your tokens, components, and UX constraints.

npm i -g @memi-design/cli
curl -fsSL https://x.ai/cli/install.sh | bash   # Grok Build

cd your-repo
memi agent install grok-build --project .
grok inspect
grok mcp doctor memoire
memi agent brief . --intent "Audit this UI" --agent grok-build --json

What memi agent install grok-build writes: