cd /news/ai-agents/brandrail-make-claude-code-and-curso… · home › topics › ai-agents › article
[ARTICLE · art-168777] src=github.com ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

Brandrail – make Claude Code and Cursor follow your design system

Brandrail launched an MCP server and npm CLI that feed a company's Brand Kit and component library to AI coding tools including Claude Code, Cursor, Copilot, Codex, Windsurf and Gemini CLI before those tools write UI, then validate the generated code and return fixes. The server exposes seven read-only tools — get_guidelines, get_design_tokens, find_component, get_component_docs, get_template, get_migration_map and validate_snippet — and runs locally over stdio, with code checked on the user's machine and never uploaded. Developers connect a repo with `npx brandrail connect`, which writes MCP config and rule files containing no key, or add the server manually via npx without an account; the same package also runs as a CLI checker with `check --changed`, `scan` and `migrate` commands.

read3 min views3 publishedNov 13, 2026
Brandrail – make Claude Code and Cursor follow your design system
Image: Michielbdejong (auto-discovered)

Your design system, in every AI coding tool.

AI coding tools don't know your design system, so every screen comes out with different colours, buttons and spacing. Brandrail's MCP server gives the tool your Brand Kit and components before it writes UI, then checks what it wrote and returns the fixes, so it corrects itself before you see the code.

  • Seven read-only tools: get_guidelines ,get_design_tokens ,find_component ,get_component_docs ,get_template ,get_migration_map ,validate_snippet .
  • Works in React, Angular, Vue, Web Components and plain HTML. The checker also reads Svelte, CSS, PHP, Twig, Mustache, Handlebars, Nunjucks and Liquid.
  • Runs locally over stdio. Code is checked on your machine and never uploaded.
  • Try it with no account: it starts with starter tokens. Connect a free Brandrail account to use your own Brand Kit.

In each repo:

npx brandrail connect

It signs you in through the browser and writes the MCP config and rule files for Claude Code, Cursor, Copilot, Codex, Windsurf and Gemini CLI. The files it writes hold no key, so commit them.

You type the code it shows on the Brandrail page, so nobody can approve a terminal for you with a link. The key is made on your machine (only its hash is sent) and saved outside the repo. To help you recognise the request, the terminal's description (user@computer · folder) is shown on the approval page.

Or add the MCP server by hand (works without an account):

{
  "mcpServers": {
    "brandrail": { "command": "npx", "args": ["-y", "brandrail", "mcp"] }
  }
}

Cursor install link:

cursor://anysphere.cursor-deeplink/mcp/install?name=brandrail&config=eyJjb21tYW5kIjoibnB4IiwiYXJncyI6WyIteSIsImJyYW5kcmFpbCIsIm1jcCJdfQ==

Set BRANDRAIL_KEY (an API key from Dashboard → Connect) to load your organisation's Brand Kit without running connect. Full docs: https://www.brandrail.com.au/docs/connect

BRANDRAIL_KEY is only ever sent to www.brandrail.com.au (or localhost). If BRANDRAIL_API_URL points anywhere else, the CLI says so on every run and won't send the key; sign in to that server with connect.

Tool What it does
get_guidelines The project's design system rules: components to use in its framework, allowed tokens, what the checker enforces. Called once before writing UI.
get_design_tokens Every colour, spacing, radius and type token with the CSS variable to use.
find_component Finds the component for a need ("date picker", "primary action") with its import, rules and an example.
get_component_docs Full docs for one component in the project's framework: import or tag, props, rules, example.
get_template A whole page (list, detail, form, settings, dashboard, sign-in) built from your components, with the files to create and the page's rules.
get_migration_map What replaces each piece of Material, MUI, Bootstrap, Ant, Chakra or PrimeNG.
validate_snippet Runs the checker on generated code and returns a brand score and each fix.

The same package is a command-line checker:

npx brandrail check --changed          # only what you changed vs origin/main
npx brandrail scan http://localhost:3000 --pages 3   # a running app, in a real browser
npx brandrail migrate                  # what's left of Material, MUI, Bootstrap and others

Every command and option: packages/cli/README.md and https://www.brandrail.com.au/docs/cli

The MCP server and CLI run on your machine. Code is checked locally and never uploaded. With a key, they fetch your Brand Kit; --upload and --report send scores and messages to your dashboard, never code. Security details: https://www.brandrail.com.au/security

packages/
  cli/      the `brandrail` npm package: CLI and local MCP server
  mcp/      the MCP server's tools
  engine/   the checker rules engine (pure TypeScript, no runtime dependencies)
mcpb/       MCP bundle for Smithery and desktop apps
plugins/
  cursor/   Cursor plugin: MCP server plus a rule
pnpm install
pnpm test     # engine tests
pnpm build    # → packages/cli/dist/brandrail.mjs

MIT, see LICENSE. The Brandrail web app (Brand Kit hosting, repo sync, GitHub App, teams) is a separate service: https://www.brandrail.com.au/pricing

Questions: hello@brandrail.com.au · Security reports: security@brandrail.com.au

── more in #ai-agents 4 stories · sorted by recency
── more on @brandrail 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/brandrail-make-claud…] indexed:0 read:3min 2026-11-13 · —