eds-mcp-server
Enables AI agents to access design system tokens and component contracts through MCP, reducing token usage and ensuring consistency.
README
eds-mcp-server — the Edwson Design System over MCP
A read-only Model Context Protocol server that hands your design-system token + component contract to any AI agent (Claude Code, Codex, Cursor, …) as callable tools — so the whole org builds UI from one source of truth, the system auto-syncs, and your AI-token bill goes down instead of up.
This is the runnable backbone behind the "AI-Native Development" section of the design system showcase.
Why an enterprise adopts this (and doesn't leave)
1 · It cuts your AI-token spend. The expensive way teams use AI for UI today: paste a 6–15 KB CSS/theme file plus a screenshot or two into the agent every turn, let it regenerate full hex/CSS, and correct it 2–3 times. That's ~10K+ input tokens × several loops per component.
This server instead serves targeted, reference-based, cache-friendly slices:
| Mechanism | Effect on tokens |
|---|---|
Targeted retrieval — get_tokens(color, dark) returns only what's asked |
a color slice is 78% smaller than the full token file (measured) |
Reference output — agent emits --accent / OrderTicket, not regenerated CSS |
output tokens drop sharply |
Contract correctness — whenToUse / props / regulatory / dataContract make it right the first time |
correction loops (the real cost driver) collapse toward 1 |
| Deterministic + compact — same input → same small output | the static slice sits behind prompt caching |
2 · It auto-syncs across the org.
build-manifest.js hashes every contract file into manifest.json (version + per-file SHA-256).
Consumers compare their cached manifest and pull only the changed files (diff_since). A token
change propagates everywhere — apps (build), Storybook (docs), agents (MCP) — without a rebuild,
because everything references tokens, never literals.
3 · It enforces consistency + compliance. One signed manifest = one source of truth; checksums prevent drift; SemVer gates breaking changes. Regulated components (risk warnings, KYC gates, sign-off bars) carry their rule in the contract and are marked non-removable, so neither a human in a hurry nor an agent can ship the surface without them.
Tools (all read-only)
| Tool | Returns |
|---|---|
list_token_groups |
group names only (cheapest orientation) |
get_tokens {group, theme?} |
only the requested token group |
search_components {query} |
matching ids + one-line purpose |
get_component {id} |
one component's full contract |
get_data_contract {id} |
data shape + required render states |
get_manifest |
version + per-file checksums |
diff_since {version} |
changed files since a version (auto-sync delta) |
Run
npm install
npm run build:manifest # regenerate manifest.json after any token/component change
npm start # speaks MCP over stdio
Register with an MCP client (example for Claude Code / Cursor mcp.json):
{ "mcpServers": { "eds": { "command": "node", "args": ["server.js"], "cwd": "./eds-mcp" } } }
Requires Node 18+. The contract lives in tokens.json + components.json; edit those, re-run
build:manifest, and every consumer picks up the delta.
Files
tokens.json token contract (color light/dark, space, radius, type, density)
components.json component contracts (purpose, when-to-use/not, props, a11y, regulatory, dataContract)
build-manifest.js auto-sync engine — hashes contracts -> manifest.json
server.js MCP server — read-only tools over stdio
manifest.json generated — version + SHA-256 per file
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.