MPDS-MCP

MPDS-MCP

A multi-project design system MCP server that manages design tokens, component specs, and validation with parent-child inheritance, accessible via REST API and JSON-RPC.

Category
Visit Server

README

MPDS-MCP

Multi-Project Design System MCP Server — an HTTP server that exposes design tokens, component specs, and validation utilities via a JSON REST API and an MCP JSON-RPC endpoint. Multiple design-system projects can coexist, each with parent-child inheritance for token resolution.

Architecture

modules/
  mcp-server/   # Express HTTP server (this binary, port MCP_PORT)
  registry/     # Project CRUD (SQLite-backed)
  tokens/       # Token storage, overrides, semantic resolution
  components/   # Component spec storage and overrides
  patterns/     # Pattern library: patterns, variants, composition rules, layout guidelines
  validate/     # Color-pair + snippet validation
  preview/      # HTML showcase generator (generate_showcase MCP tool)
  db/           # Shared SQLite connection + migrations

Running locally

Prerequisites

  • Node.js 20+ and npm
  • A writable directory for the SQLite DB (or use :memory: for a transient session)

Start (development)

npm ci
MCP_SECRET=dev DB_PATH=/tmp/mpds-dev.db MCP_PORT=3100 \
  npm --prefix modules/mcp-server run dev

Health check:

curl http://localhost:3100/health
# → {"status":"ok"}

Start (production build)

npm ci
npm --prefix modules/mcp-server run build
MCP_SECRET=<secret> DB_PATH=/home/mpds/data/mpds.db MPDS_ENV=production \
  npm --prefix modules/mcp-server start

Environment variables

Variable Required Default Description
MCP_SECRET Yes (production) "" (no auth) Bearer token — every /api/* request must carry it
DB_PATH Yes SQLite file path or :memory: (transient)
MCP_PORT No random port TCP port the server listens on
MPDS_ENV No Set to production to enforce MCP_SECRET check
HOME No (set by OS/Docker) Used for allowed DB path prefix validation

API endpoints

All endpoints (except /health) require Authorization: Bearer <MCP_SECRET>.

Error envelope: { "error": { "code": "...", "message": "..." } }

Health

GET /health

Projects

GET    /api/projects
POST   /api/projects           body: { id, name, parentId? }
GET    /api/projects/:id
DELETE /api/projects/:id

Tokens

GET    /api/projects/:projectId/tokens
PUT    /api/projects/:projectId/tokens/:key     body: { value, description? }
DELETE /api/projects/:projectId/tokens/:key/override

Components

GET    /api/projects/:projectId/components
GET    /api/projects/:projectId/components/:componentId
PUT    /api/projects/:projectId/components/:componentId/override
DELETE /api/projects/:projectId/components/:componentId/override

Validation

POST   /api/validate/color-pair     body: { foreground, background }

Patterns

GET    /api/projects/:projectId/patterns
POST   /api/projects/:projectId/patterns                       body: { id, name, category, description?, tags?, guidanceUrl? }
GET    /api/projects/:projectId/patterns/:patternId
PATCH  /api/projects/:projectId/patterns/:patternId            body: { name?, description?, tags?, guidanceUrl? }
DELETE /api/projects/:projectId/patterns/:patternId

Pattern variants

GET    /api/projects/:projectId/patterns/:patternId/variants
POST   /api/projects/:projectId/patterns/:patternId/variants   body: { name, appliesAt, description? }
GET    /api/projects/:projectId/patterns/:patternId/variants/:variantId
PATCH  /api/projects/:projectId/patterns/:patternId/variants/:variantId
DELETE /api/projects/:projectId/patterns/:patternId/variants/:variantId

Composition rules

GET    /api/projects/:projectId/composition-rules
POST   /api/projects/:projectId/composition-rules              body: { patternAId, patternBId, relation, guidance? }
DELETE /api/projects/:projectId/composition-rules/:ruleId

relation enum: NESTING_ALLOWED | NESTING_FORBIDDEN | OVERRIDE_CAUTION | SIBLING_ONLY | EXCLUSIVE

Layout guidelines

GET    /api/projects/:projectId/layout-guidelines
POST   /api/projects/:projectId/layout-guidelines              body: { type, name, description?, data }
GET    /api/projects/:projectId/layout-guidelines/:guidelineId
PATCH  /api/projects/:projectId/layout-guidelines/:guidelineId
DELETE /api/projects/:projectId/layout-guidelines/:guidelineId

type enum: breakpoints | spacing | grid | alignment | typography | animation

See contracts/P1/ and contracts/P2/ for the full frozen OpenAPI specs.

MCP configuration

The server exposes a JSON-RPC 2.0 endpoint at POST /mcp (requires the same Authorization: Bearer <MCP_SECRET> header as the REST API). Configure it in your MCP client as an HTTP server.

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json)

{
  "mcpServers": {
    "mpds": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "http://localhost:3100/mcp"],
      "env": {
        "MCP_REMOTE_HEADER_AUTHORIZATION": "Bearer <MCP_SECRET>"
      }
    }
  }
}

Claude Code (.claude/settings.json in your project)

{
  "mcpServers": {
    "mpds": {
      "type": "http",
      "url": "http://localhost:3100/mcp",
      "headers": {
        "Authorization": "Bearer <MCP_SECRET>"
      }
    }
  }
}

GitHub Copilot (.vscode/mcp.json in your workspace)

{
  "inputs": [
    {
      "type": "promptString",
      "id": "mpds_secret",
      "description": "MPDS MCP bearer token",
      "password": true
    }
  ],
  "servers": {
    "mpds": {
      "type": "http",
      "url": "http://localhost:3100/mcp",
      "headers": {
        "Authorization": "Bearer ${input:mpds_secret}"
      }
    }
  }
}

The input block causes VS Code to prompt for the secret once per session and store it in the system keychain. To skip the prompt, replace ${input:mpds_secret} with the token literal (not recommended for shared workspaces).

OpenCode (opencode.json in your project root, or ~/.config/opencode/opencode.json globally)

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "mpds": {
      "type": "remote",
      "url": "http://localhost:3100/mcp",
      "enabled": true,
      "headers": {
        "Authorization": "Bearer <MCP_SECRET>"
      }
    }
  }
}

To avoid hard-coding the secret, reference an environment variable using OpenCode's {env:VAR} interpolation:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "mpds": {
      "type": "remote",
      "url": "http://localhost:3100/mcp",
      "enabled": true,
      "headers": {
        "Authorization": "Bearer {env:MCP_SECRET}"
      }
    }
  }
}

Available MCP methods

Read & validate

Method Description Required params
list_projects List all design-system projects
get_tokens Resolve tokens for a project (with inheritance) projectId, category?
get_design_system Full token map + component specs projectId
get_component_spec Single component spec projectId, componentId
validate_color_pair WCAG contrast ratio fg, bg, context (normal|large|ui)
validate_token_pair Contrast check using token keys projectId, fgKey, bgKey, context
validate_snippet Lint HTML/JSX for a11y issues content, contentType? (html|jsx)
create_guideline Create a design guideline id, title, body?, tags?
search_guidelines Full-text search guidelines query
propose_token_override Propose a token value change for review projectId, tokenKey, proposedValue, rationale?, agentId?
list_proposals List pending token proposals projectId

Write — projects & tokens

Method Description Required params
create_project Create a project id, name, parentId?
update_project Rename a project projectId, name
delete_project Delete a project projectId
list_tokens List all tokens for a project projectId, category?
get_token Get a single token projectId, key
create_token Create a token projectId, key, category, value, isSemantic?, semanticRef?
update_token Update a token value (OCC) projectId, key, version, value?, semanticRef?
set_token Set/override a token value projectId, key, value, version
delete_token Delete a token (OCC) projectId, key, version
delete_token_override Remove a child project override projectId, key

Write — components

Method Description Required params
create_component Create a component spec projectId, id, name, description?, props?, variants?, states?, …
update_component Update a component spec (OCC) projectId, componentId, version, name?, description?, …
delete_component Delete a component spec projectId, componentId

Write — pattern library

Method Description Required params
create_pattern Create a pattern projectId, id, name, category, description?, tags?
update_pattern Update a pattern projectId, patternId, name?, description?, tags?
delete_pattern Delete a pattern projectId, patternId
create_variant Add a variant to a pattern projectId, patternId, name, appliesAt, description?
update_variant Update a variant projectId, patternId, variantId, name?, appliesAt?
delete_variant Delete a variant projectId, patternId, variantId
create_composition_rule Define a pattern relationship projectId, patternAId, patternBId, relation, guidance?
delete_composition_rule Remove a composition rule projectId, ruleId
create_layout_guideline Create a layout guideline projectId, type, name, description?, data
update_layout_guideline Update a layout guideline projectId, guidelineId, name?, description?, data?
delete_layout_guideline Delete a layout guideline projectId, guidelineId

Showcase

Method Description Required params
generate_showcase Generate a self-contained HTML design system preview (color palette, component gallery, pattern library) projectId, title?
# Save and open the showcase locally
curl -s http://localhost:3100/mcp \
  -H "Authorization: Bearer <MCP_SECRET>" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"generate_showcase","params":{"projectId":"my-ds"}}' \
  | jq -r '.result.html' > /tmp/showcase.html && open /tmp/showcase.html

All requests follow JSON-RPC 2.0:

curl -s http://localhost:3100/mcp \
  -H "Authorization: Bearer <MCP_SECRET>" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"list_projects","params":{}}' | jq

Running tests

Tests use an in-memory SQLite database and require no external services:

DB_PATH=:memory: MCP_SECRET=test npm test

Docker

docker build -t mpds-mcp .
docker run -p 3100:3100 \
  -e MCP_SECRET=<secret> \
  -e DB_PATH=/home/mpds/data/mpds.db \
  -v $(pwd)/data:/home/mpds/data \
  mpds-mcp

See INSTALL.md for full setup and docker-compose instructions.

Recommended Servers

playwright-mcp

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.

Official
Featured
TypeScript
Magic Component Platform (MCP)

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.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

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.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

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.

Official
Featured
TypeScript
Kagi MCP Server

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.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

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.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured