mcpField
C
Carbon MCP

Carbon MCP

Remote
github.com/carbon-design-systemDeveloper Tools41Updated 2w ago
Repository

About Carbon MCP

Carbon MCP is a hosted server for teams using AI agents to work with the Carbon Design System. Its tools provide searchable design documentation and code examples, chart samples, and checks against Carbon guidelines.

Key features

  1. 01Search Carbon component code examples
  2. 02Find design system documentation
  3. 03Retrieve chart code examples
  4. 04Search Carbon Labs components
  5. 05Audit code against Carbon guidelines

Use cases

  1. 01Generate Carbon-based interface code
  2. 02Check components and tokens in code
  3. 03Look up accessibility and usage guidance
  4. 04Find chart examples for a framework

MCP support

The hosted MCP server exposes Carbon search, chart, and code-audit tools to MCP clients. Clients connect to its remote endpoint and authenticate with IBMid.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.carbondesignsystem.com/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

Carbon MCP: the Carbon Design System, available to any AI agent. Components, icons, charts and documentation served to AI clients from one hosted MCP server.

Carbon MCP

The Carbon Design System, available to any AI agent.

Carbon MCP is a hosted server that gives AI agents direct access to Carbon's components, icons, charts and documentation through MCP, the open Model Context Protocol standard for connecting AI applications to external tools and data. It runs on IBM Cloud, so any MCP client can connect with a URL and an IBMid sign-in. This repository is where you report issues, request features and get the Carbon Builder skill.

Connect · Carbon Builder skill · Tools · Coverage · Report an issue

Terms of Use for the hosted service · Apache-2.0 for the Carbon Builder skill


Why Carbon MCP

Carbon MCP makes any agentic AI application an instant expert in the Carbon Design System, so teams deliver consistent, high-quality user experiences with greater speed.

BenefitWhat it means
Direct access to CarbonAI agents query the Carbon Design System for the context they need: colors, typography, component source, usage guidelines.
Better generated codeCode examples and documentation from Carbon MCP let AI tools generate high-fidelity Carbon UI code that follows design and development practice.
One answer for the whole teamDesigners, developers and PMs get consistent answers and examples from a shared source of truth, which reduces drift and rework.

Connect

The Carbon MCP server is hosted remotely at https://mcp.carbondesignsystem.com/mcp. Add it to your client as a remote MCP server and sign in with your IBMid when asked.

Clients that read an mcp.json file use this minimal configuration:

{
  "mcpServers": {
    "carbon-mcp": {
      "type": "streamable-http",
      "url": "https://mcp.carbondesignsystem.com/mcp",
      "oauth": true,
      "alwaysAllow": []
    }
  }
}
ClientSetup
IBM BobOnboarding: IBM Bob
Claude DesktopOnboarding: Claude Desktop
CursorOnboarding: Cursor
Figma MakeOnboarding: Figma Make
Claude Code, VS Code, GitHub Copilot and other clientsOnboarding: Other clients
Clients without OAuthToken and session ID

The full onboarding and setup guide covers what to have before you begin, the prerequisites, connecting to the server and every client above.

Token and session ID

If your client does not support OAuth discovery, you can authenticate with a token and session ID instead.

  1. Generate a token and session ID.
  2. Copy your credentials and add them to your Carbon MCP configuration.

Pair it with the Carbon Builder skill

Carbon Builder is a portable AI skill that turns an agent into an expert Carbon Design System engineer. It retrieves live component examples, variants, props and documentation from the Carbon MCP server, generates Carbon React and Web Components code, and builds Carbon Charts across every supported framework.

The server supplies the facts; the skill tells the agent when to ask for them and how to turn the answers into working code. Together they produce better results than either alone.

Install it with the skills CLI, choosing the line for your agent:

npx skills add https://github.com/carbon-design-system/carbon-mcp/tree/main/public/skills/carbon-builder -g -a bob
npx skills add https://github.com/carbon-design-system/carbon-mcp/tree/main/public/skills/carbon-builder -g -a claude-code
npx skills add https://github.com/carbon-design-system/carbon-mcp/tree/main/public/skills/carbon-builder -g -a cursor
npx skills add https://github.com/carbon-design-system/carbon-mcp/tree/main/public/skills/carbon-builder -g -a codex
npx skills add https://github.com/carbon-design-system/carbon-mcp/tree/main/public/skills/carbon-builder -g -a github-copilot
npx skills add https://github.com/carbon-design-system/carbon-mcp/tree/main/public/skills/carbon-builder -g -a gemini-cli

The CLI supports many more agents; see skills.sh. The skill has its own page at skills.sh/carbon-design-system/carbon-mcp/carbon-builder.

To install without the CLI, download carbon-builder.zip, unzip it, and copy the carbon-builder/ folder to your client's skills location. The skill README lists the path for each client, and the onboarding guide walks through it in Set up your AI client and install the Carbon Builder skill.

Tools

ToolWhat it returns
code_auditA check of your code against Carbon guidelines covering design tokens, component usage, accessibility, layout and best practices.
code_searchCarbon React and Web Components code examples, icons and pictograms, including complete example application files.
docs_searchCarbon Design System and IBM Products documentation, such as component guidance, usage, accessibility and reference material.
get_chartsCarbon Charts code examples for React, Angular, Vue, Svelte, Vanilla JS and HTML.
labs_searchCode examples, props, events and usage docs for Carbon Labs experimental components.

Coverage

Carbon MCP continues to evolve. Current coverage:

CorpusExamplesWhat is indexed
@carbon/reactButton, Tabs, TextInputCode examples and documentation, multiple variants per component, complete props and imports
@carbon/web-componentsbutton, tabs, text-inputCode examples and documentation, multiple variants per component, complete props and imports
Carbon for IBM ProductsTearsheets, PageHeaderCode examples and documentation, multiple variants per component, complete props and imports
Carbon Icons and PictogramsSettings, User, UserAvatarFilledCode examples and documentation, usage
@carbon/ai-chatbasic, custom-element, historyCode examples and documentation, full code examples, complete props and imports
@carbon/chartsChart types, with links to demosChart type descriptions, links to demos, full code examples, complete props and imports
@carbon-labs/*React: AnimatedHeader, Processing, Resizer, UIShell, What's NewCode examples and documentation for preview candidates, multiple variants per component, complete props and imports

The Concepts page describes the content in more detail.

Documentation

DocumentCovers
OverviewCarbon MCP overview and key benefits
ConceptsThe Carbon MCP tools and content coverage
Onboarding and setupAuthentication, connecting common MCP clients, installing the Carbon Builder skill
PromptsPrompting guidance and best practices
Token conservationKeeping AI workflows fast and cost-efficient by cutting unnecessary context and responses
Terms of UseTerms for the IBM Cloud-hosted Carbon MCP service
SupportGetting help
Code of conductHow we expect everyone to behave in this project

Issues and support

Open an issue to report a bug, request a feature or ask for help. Templates guide each one.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.