Brandsystem MCP — Build

What Brandsystem MCP — Build MCP server does

Brandsystem MCP — Build converts existing brand material into structured files that AI agents and design workflows can use. Sources include a website URL, PDF guidelines, a Figma library, local candidates, or a hosted Brandcode Studio brand. The resulting .brand/ directory can contain core identity data, DTCG design tokens, a single-file brand-runtime.json contract, interaction policies, design synthesis data, an agent-readable design brief, extracted logo assets, and an offline HTML report.

The server covers the full adopt, use, and check workflow. It can discover or extract colors, fonts, and logos; compile those findings into reusable brand data; ask for confirmation when values are ambiguous; and produce task-specific context for writing, design, or implementation work. It also supports exporting the compiled system for AI chats, code projects, teams, email, PDFs, and Claude skills.

How it works

Start with brand_start, optionally providing a website, guideline PDF, Figma file key, or Brandcode Studio URL. Automatic website mode can run extraction, compilation, report generation, and runtime creation in one operation. brand_status shows progress and identifies the next action when work is incomplete.

After extraction, brand_compile creates tokens and runtime files. brand_clarify resolves uncertain values such as color roles or font names. Use brand_runtime for the complete contract or a smaller visual, voice, or minimal slice. For task-specific work, brand_context selects governed rules and returns the matching selectors and output contract rather than passing the entire runtime.

Validation tools serve different stages. brand_check performs fast inline checks on text, colors, fonts, or CSS while content is being created. brand_preflight checks HTML or CSS for broader rule compliance, including off-brand colors, incorrect fonts, missing logos, and listed anti-patterns. brand_report creates a self-contained HTML report suitable for offline viewing or upload to an AI chat.

Setup and configuration

Install the npm package with an MCP client installation command or add it to an MCP configuration using npx:

{
  "mcpServers": {
    "brandsystem": {
      "command": "npx",
      "args": ["-y", "@brandsystem/mcp"]
    }
  }
}

The package provides installation targets for Codex, Claude Code, Cursor, Windsurf, Claude Desktop, and Cline. The installer performs a dry run unless --write is supplied. Claude Desktop also supports a downloadable MCP bundle from the project's latest GitHub release, which asks for the folder containing the .brand/ directory.

Authentication is optional for extraction, preview, and local checking. Call brand_brandcode_auth when saving or pushing a brand to Brandcode Studio, then use brand_brandcode_connect to pull a hosted brand or upload the local runtime. Local operation does not require an account.

Tools and capabilities

• Adopt brand data from websites, PDFs, Figma, local candidates, or Brandcode Studio.
• Compile identity findings into DTCG tokens, runtime contracts, policies, and design artifacts.
• Select compact, task-specific context for agents and sub-agents.
• Check text, typography, colors, HTML, and CSS against brand rules.
• Generate offline reports and destination-specific exports, including chat, code, team, email, PDF, and Claude skill formats.
• Authenticate with Brandcode Studio and pull or save shared brand systems.