grocers-design-mcp
Exposes the Grocers Design System as queryable tools for coding agents, including design tokens, UI components, brand foundations, and a skill manifest.
README
grocers-design-mcp
MCP server exposing the Grocers Design System as queryable tools for coding agents.
Hosted on Fly.io. Auth is a single long-lived token embedded in the URL path — no OAuth, no client config beyond pasting the URL.
Live: https://grocers-design-mcp.fly.dev/<TOKEN>/mcp
Health (no auth): https://grocers-design-mcp.fly.dev/health
What it serves
- Design tokens — every CSS custom property from the Grocers DS (colors, spacing, radii, shadows, type scale, motion). 101 tokens indexed at boot.
- Semantic type classes —
t-h1,t-display,t-body,t-price, etc. (16 ready-to-apply utility classes). - UI components — JSX source from the
web,mobile, andadminUI kits. 44 components — Dashboard, KpiCard, OrdersScreen, OrderRow, CartSheet, MProductTile, etc. - Brand foundations — content rules (tone, casing, no-emoji), visual rules (red ≠ error, destructive uses ink + type-to-confirm), iconography (Lucide, 2px stroke, 24px grid).
- Skill manifest — one-paragraph brand brief for fast agent bootstrap.
Tools
| Tool | Purpose |
|---|---|
list_tokens({category?, group?}) |
Discover tokens by category/group |
get_token({name}) |
Full detail for one token (with -- prefix) |
list_semantic_classes() |
t-* pre-built typography classes |
list_components({kit?}) |
Components by kit (web/mobile/admin) |
get_component({name, kit?}) |
JSX source + metadata |
get_foundations({section?}) |
README sections: visual/content/iconography/… |
get_skill() |
The SKILL.md manifest |
search({query, limit?}) |
Full-text search with pt-BR↔en synonyms |
validate_snippet({snippet, language?}) |
Lint CSS/JSX against 11 brand rules |
suggest_component({intent, kit?, include_source?}) |
Intent → idiomatic component or pattern |
Lint rules in validate_snippet
| Rule | Severity | Catches |
|---|---|---|
no-emoji |
error | Any emoji in source (incl. composed sequences) |
no-hardcoded-brand-color |
warning | #ff4444, #ed1f1f, … → suggests exact token |
no-hardcoded-ink-color |
warning | #0b0119, #14081f, … → suggests exact token |
no-arbitrary-tailwind |
warning | bg-[#ff4444], p-[13px], rounded-[7px] |
red-is-not-error |
warning | Brand red near 'erro/error/invalid' → use danger |
destructive-uses-ink-not-red |
warning | Destructive cue + brand red → ink + type-to-confirm |
off-scale-radius |
warning | border-radius not in {4,6,10,14,18,24,pill} |
off-scale-spacing |
warning | padding/margin/gap not in 4px scale |
non-brand-font |
warning | font-family not Neue Montreal/Geist Mono |
heavy-cta-weight |
info | font-weight: 600+ in button context |
missing-tabular-nums |
info | R$ N without tabular-nums or .t-price |
off-scale-shadow |
info | box-shadow not via var(--shadow-*) |
Intent map in suggest_component
30 curated entries spanning all three kits plus 5 non-component patterns
(destructive-action, error-state, price-display, icon-usage,
tone-of-voice). Kit inference from natural language ("mobile", "painel",
"web") biases the ranking.
Using with Claude Code
Add to ~/.claude.json or project .mcp.json:
{
"mcpServers": {
"grocers-design": {
"type": "http",
"url": "https://grocers-design-mcp.fly.dev/<TOKEN>/mcp"
}
}
}
The token lives in fly secrets and rotates via
fly secrets set GROCERS_MCP_TOKEN=$(openssl rand -hex 32) -a grocers-design-mcp.
Quick example flow (agent building a Grocers admin screen)
1. get_foundations({section:"visual"}) → brand rules
2. suggest_component({intent:"tabela de pedidos no admin"})
→ admin/OrdersScreen
3. suggest_component({intent:"modal de cancelar pedido"})
→ pattern destructive-action
(ink + type-to-confirm)
4. get_component({name:"OrdersScreen", kit:"admin"})
→ JSX source
5. validate_snippet({snippet: <composed JSX>}) → catch drift before ship
Local dev
npm install
GROCERS_MCP_TOKEN="$(openssl rand -hex 32)" PORT=8080 npm run dev
# Health: http://localhost:8080/health
# MCP: http://localhost:8080/<TOKEN>/mcp
Architecture
- Hono +
@hono/node-serverfor HTTP. - @modelcontextprotocol/sdk for JSON-RPC / streamable HTTP transport,
stateless mode (fresh
McpServerper request — no session state). - Token-in-URL middleware:
timingSafeEqualagainstGROCERS_MCP_TOKEN; wrong token returns 404 (not 401 — hides existence). - The DS bundle ships inside the container at
bundle/and is indexed once at boot into in-memory maps (O(1) lookups). - Fly app runs 2 machines (HA) in
gru, 256MB each, auto-stops when idle.
Deploy
fly deploy -a grocers-design-mcp --remote-only
Layout
bundle/ # Grocers DS handoff bundle (source of truth)
project/
colors_and_type.css # all design tokens
README.md # foundations
SKILL.md # one-paragraph brand brief
ui_kits/{web,mobile,admin}/ # React/JSX components
preview/*.html # visual specimens (not served by MCP)
src/
index.ts # entrypoint
server.ts # Hono app + token middleware
indexer.ts # CSS + JSX + README parser
tools/
tokens.ts # list_tokens, get_token, list_semantic_classes
components.ts # list_components, get_component
foundations.ts # get_foundations, get_skill
search.ts # search (synonym-aware)
validate.ts # validate_snippet (11 rules)
suggest.ts # suggest_component (30 entries)
Dockerfile
fly.toml
Roadmap
Out of scope for v0.3, candidates for next iteration:
- Multi-theme support (Zona Sul and other clients as override layers).
validate_snippet: semantic copy lint (e.g. "Erro: ..." string detection).suggest_component: broader UI primitive coverage (TableHeader, Pagination states, EmptyState variants).get_screen(name): return a parent component + all its children in one call.- shadcn/ui compatibility layer (map Grocers semantic tokens →
--primary/--backgroundetc.).
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.