@sorb/tap

@sorb/tap

MCP server for design tokens that gives AI agents read access to design systems, enabling token resolution, listing, and component usage lookups via natural language.

Category
Visit Server

README

<p align="center"> <img src="docs/banner.svg" alt="@sorb/tap — design tokens on tap for AI agents; an MCP server for the Sorb design system" width="100%"> </p>

@sorb/tap

Give your AI agent read access to your design system. @sorb/tap is an MCP (Model Context Protocol) server for design tokens: point it at a DTCG token file, a Sorb™ project, or a running Sorb bridge, and your agent can answer questions like "where does this button color come from?" — with the actual alias chain and the components that use it. Read-only by design.

Quickstart

# Claude Code — try it against the bundled example set:
claude mcp add sorb-tap -- npx -y @sorb/tap --tokens node_modules/@sorb/tap/examples/acme-tokens.json

# …or your own DTCG file(s):
claude mcp add sorb-tap -- npx -y @sorb/tap --tokens tokens.json

Then ask: "What's the button background token, and where does it come from?"

demo: Claude resolving a token through its alias chain

Other MCP clients (Cursor, LangChain, anything that speaks MCP over stdio): register the same command — npx -y @sorb/tap --tokens tokens.json — as a stdio server, using your client's own config format.

What the agent sees

Real output against examples/acme-tokens.json:

▸ resolve_token { "id": "component.button.bg" }
{
  "id": "component.button.bg",
  "cssVar": "--component-button-bg",
  "value": "#3b82f6",
  "chain": ["semantic.brand.primary", "primitive.blue.500"]
}

▸ list_tokens { "tier": "semantic" }
{
  "count": 4,
  "tokens": [
    { "id": "semantic.brand.primary", "cssVar": "--semantic-brand-primary",
      "value": "#3b82f6", "tier": "semantic", "type": "color" },
    …
  ]
}

Tools

Six tools, all read-only. Token ids are dot paths (component.button.bg); everywhere an id is accepted, the CSS variable form (--component-button-bg) works too.

tool what it returns
list_tokens(tier?, type?, q?) tokens with id, CSS variable, resolved value, tier, type — filter by tier (primitive/semantic/component), $type, or substring
get_token(id) one token, raw authored value + resolved
resolve_token(id) final value with the alias chain that produced it
list_components() captured components/stories
get_component_bindings(component) role → token map (fill/stroke/cornerRadius/…)
find_token_usage(id) reverse index: which components bind a token, and in which roles

Plus one resource, sorb-tap://tokens — the full resolved token map as JSON.

The three component tools need Sorb capture artifacts (sorb-seed capture) or a bridge; in bare-DTCG mode they say so instead of returning silent emptiness, so the agent knows what would unlock them.

Sources

flag serves
--tokens <file.json> … bare DTCG token file(s) — no Sorb required
--dir <project> a Sorb project: .sorb/resolved.json + capture artifacts
--bridge <url> [--pk <key>] a running Sorb bridge (sorb dev); hosted bridges — e.g. bridge.sorbcloud.com — work with a read-only sorb_pk_ key from your workspace

No flags: auto-detects .sorb/ in the cwd, else tokens/*.json. Env: SORB_TAP_BRIDGE (bridge URL), SORB_TAP_PK (read-only key).

LangChain

A runnable example agent lives in examples/langchain-client — the same six tools loaded through @langchain/mcp-adapters into a LangChain (LangGraph) agent.

Limitations

  • Read-only. No tool writes, proposes, or applies changes — by design.
  • The DTCG reader resolves $value/$type/{alias} references only: no Style Dictionary transforms, no math expressions, no modes/theming. For those, run the file through your build and point tap at the output — or at a Sorb project, where .sorb/resolved.json is the built map.
  • Bridge mode snapshots at startup; restart (or call refresh() when embedding) to pick up new captures.
  • Component bindings come from Sorb captures — bare DTCG files have none.

Programmatic use

import { loadDtcgFiles, runTool } from '@sorb/tap'
const source = loadDtcgFiles(['tokens.json'])
runTool(source, 'resolve_token', { id: 'component.button.bg' })
// → { id, cssVar, value: '#3b82f6', chain: ['semantic.brand.primary', …] }

Development

npm install && npm test    # node:test, 17 cases, incl. a stub bridge
npm run build              # esbuild → dist/ (JS only — no TypeScript, ever)

Roadmap: authenticated write tools (propose/apply) may arrive later via Sorb Cloud; the read surface here stays free and open.


Part of the Sorb tree — Seed captures · Juice bridges · Leaf renders · Canopy designs · Tap is how agents draw from it.

MIT © Metatoy LLC · Sorb™ is a trademark of Metatoy LLC. Works with Figma. Not affiliated with, or endorsed by, Figma. Figma is a trademark of Figma, Inc.

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