Angular Storybook MCP Server
An MCP server that brings an Angular design system built on Storybook into AI tools like Claude and Cursor, providing real-time access to component metadata, APIs, and documentation.
README
Angular Storybook MCP Server
An MCP (Model Context Protocol) server that brings an Angular design system built on Storybook into Claude Code and Cursor, enabling AI-assisted development with real-time access to component metadata, APIs, and documentation.
This server only works against a design system whose Storybook build emits the manifests/components.json and manifests/docs.json files described below — it is not a generic Angular DS integration.
Overview
This MCP server exposes an Angular Storybook design system as a set of tools available to Claude. It fetches component metadata from a live Storybook instance and provides tools for:
- Listing all components — get a complete inventory of the design system
- Searching components — find components by name or keyword
- Component imports — get correct import paths for any component
- Component props — view all props, types, and defaults for a component
- Type details — inspect complex TypeScript types used by components
- Foundations — access design tokens (colors, typography, spacing, etc.)
- Peer dependencies — pointer to where required package versions can be found (not yet in the manifests)
Before you start: two things are required
This project has two independent halves, and both must be set up — one without the other doesn't work:
- Install and configure this MCP server locally — see Setup below. This is the client-side piece that Claude Code or Cursor talks to.
- Make your Storybook emit
manifests/components.jsonandmanifests/docs.json— see Bringing up a new design system repo. This is a one-time change to the Storybook project itself (Compodoc + the manifest addon), and it must re-run on every Storybook rebuild so the manifests stay current. Without it, there's nothing for this MCP server to fetch, no matter how correctly step 1 is configured.
Setup
Both Claude Code and Cursor talk to this MCP server using the same mcpServers config shape — only the config file location differs. Pick your editor below.
Setup for Cursor
Once the Storybook owner deploys with the manifests/components.json and manifests/docs.json files included, configure Cursor to use the live URL.
Edit .cursor/mcp.json (project-level, in your repo root) or ~/.cursor/mcp.json (global, applies to all projects):
{
"mcpServers": {
"angular-ds": {
"command": "node",
"args": ["<path-to-angular-ds-mcp-server>/dist/server.js"],
"env": {
"STORYBOOK_URL": "https://your-storybook-host.example.com"
}
}
}
}
Then reload Cursor (Command Palette → "Reload Window", or fully restart Cursor). Open Cursor Settings → MCP to confirm the angular-ds server shows as connected.
Setup for Claude Code
Once the Storybook owner deploys with the manifests/components.json and manifests/docs.json files included, configure Claude Code to use the live URL.
Edit ~/.claude/claude.json:
{
"mcpServers": {
"angular-ds": {
"command": "node",
"args": ["<path-to-angular-ds-mcp-server>/dist/server.js"],
"env": {
"STORYBOOK_URL": "https://your-storybook-host.example.com"
}
}
}
}
Then restart Claude Code. The tools will be available in all sessions.
Testing
After configuring Claude Code or Cursor, start a new chat and ask:
List all components in the design system
The assistant should call list_components and return the full component list. If it works, the MCP server is properly configured.
Building
npm install
npm run build
The built server will be at dist/server.js.
Development
Run the server in dev mode (with hot reload via tsx):
npm run dev
Or start the built server directly:
npm start
Architecture
src/server.ts— Main MCP server entry point; registers all toolssrc/fetcher.ts— Handles fetching and cachingmanifests/components.jsonandmanifests/docs.jsonfrom Storybooksrc/tools/— Individual tool implementations:list-components.ts— List all componentssearch-components.ts— Search by name/keywordget-import.ts— Get import pathget-component-props.ts— Get component propsget-type-details.ts— Inspect TypeScript typesget-foundations.ts— Get design tokensget-peer-dependencies.ts— Get version requirements
Environment Variables
STORYBOOK_URL— Base URL where the manifest files are served. Required; the server throws on startup if it is not set.
Troubleshooting
"Failed to fetch component metadata"
- Check that
STORYBOOK_URLpoints to a valid URL - Verify
manifests/components.jsonandmanifests/docs.jsonexist at{STORYBOOK_URL}/manifests/components.jsonand{STORYBOOK_URL}/manifests/docs.json - For local testing, ensure the http-server is running on the correct port
Tools not appearing in Claude Code / Cursor
- Verify
~/.claude/claude.json(Claude Code) or.cursor/mcp.json/~/.cursor/mcp.json(Cursor) is properly formatted JSON - Check that the
dist/server.jsfile exists and is executable - Restart Claude Code, or reload/restart Cursor, after updating the config
Slow first query
- The server caches metadata on startup. First query may take a few seconds while it fetches from Storybook.
Integration with your Design System
This server depends on your Angular Design System's Storybook build including two manifest files, manifests/components.json and manifests/docs.json. These are automatically generated as part of the Storybook build process and contain:
manifests/components.json— component names, selectors, import statements, story snippets, and prop definitions (types, required flags, defaults, descriptions)manifests/docs.json— design tokens and other foundations documentation
Peer dependency information is not currently part of the manifests; the get_peer_dependencies tool points users to your design system package's own peerDependencies instead.
Bringing up a new design system repo (for Storybook owners)
None of this happens with a vanilla Storybook install — it requires two things wired into the target Angular repo:
1. Generate Angular metadata with Compodoc
Compodoc is a separate tool, not part of Storybook. Install it and add a script that runs it before Storybook starts/builds:
// package.json
"scripts": {
"compodoc:lib": "compodoc -p <path-to-your-lib-tsconfig> -e json -d <output-dir> --silent",
"prestorybook": "npm run compodoc:lib",
"prebuild-storybook": "npm run compodoc:lib"
}
This produces a documentation.json file describing every component's inputs, outputs, and types.
2. Add the manifest addon + feature flag to .storybook/main.ts
addons: [
'storybook-addon-angular-manifest', // must come BEFORE addon-docs
'@storybook/addon-docs',
// ...your other addons
'@storybook/addon-mcp',
],
features: {
componentsManifest: true, // Storybook 10.3+; use `experimentalComponentsManifest` on 10.2.x
},
Every story's meta must set component: (e.g. component: ButtonComponent) — without it, the addon can't resolve the Angular component name and the manifest entry comes out as an error.
3. Build and deploy Storybook as usual
npm run build-storybook now emits manifests/components.json and manifests/docs.json as static files alongside the rest of the Storybook site. Deploy that output wherever your Storybook is normally hosted (S3+CloudFront, a container behind nginx, etc.) — no special MCP-aware hosting is required, it's just two extra static JSON files.
4. Point this MCP server at it
Once the manifests are reachable at {your-storybook-url}/manifests/components.json and {your-storybook-url}/manifests/docs.json, set STORYBOOK_URL to that base URL (see Setup and Environment Variables above) — no changes to this server's code are needed.
Consumers on your team then just add this server to their own .cursor/mcp.json or ~/.claude/claude.json as shown in Setup, pointing STORYBOOK_URL at your deployed Storybook. Each person still builds/runs their own local copy of this server (it's a stdio server, not a hosted one) — only the Storybook + manifests need central deployment.
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.