MCP Server

The aidimension MCP server exposes every aidimension resource over the Model Context Protocol, so your AI agent can search components, copy templates, install skills, and generate themes — all without copy-pasting docs.

Install#

Claude Desktop#

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

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

Restart Claude Desktop. You should see "aidimension" listed in the MCP servers panel.

Cursor#

Open Settings → Features → Model Context Protocol. Add:

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

Or use the CLI:

aidimension mcp init cursor

Other agents#

The MCP server speaks stdio JSON-RPC 2.0, the standard MCP transport. Any agent that supports MCP can use it.

# Run manually to test
npx @aidimension/mcp@latest
# → {"jsonrpc":"2.0","result":{...}}

Tools#

The server exposes 12 tools:

search_components#

Search the component registry by name, prop, or use case.

mcp.aidimension.search_components({
  query: "modal dialog with form",
  limit: 5
})
// → [{ name: "Dialog", score: 0.92, ... }]

get_component#

Fetch the full source and API reference for a single component.

mcp.aidimension.get_component({ name: "Dialog" })
// → { name, source, props, examples, ... }

copy_component#

Write a component's source into the current project.

mcp.aidimension.copy_component({
  name: "Button",
  destination: "src/components/ui/button.tsx"
})

list_templates#

List available page templates.

mcp.aidimension.list_templates({ category: "marketing" })

copy_template#

Scaffold a template into the project.

mcp.aidimension.copy_template({
  name: "pricing",
  destination: "src/app/pricing/page.tsx"
})

get_theme#

Fetch the current theme tokens (colors, radii, fonts).

set_theme_token#

Update a single design token.

mcp.aidimension.set_theme_token({
  token: "--color-primary",
  value: "#7c3aed"
})

install_skill#

Install an AI skill into the current agent.

mcp.aidimension.install_skill({ name: "aidimension-component" })

list_skills#

List all available skills, with their descriptions and version.

magic_build#

Generate UI from a natural-language prompt.

mcp.aidimension.magic_build({
  prompt: "a pricing page with 3 tiers and a monthly/annual toggle",
  destination: "src/app/pricing/page.tsx"
})

search_docs#

Full-text search across the entire aidimension documentation.

mcp.aidimension.search_docs({ query: "dark mode", limit: 3 })

run_cli#

Execute any CLI command from the agent.

mcp.aidimension.run_cli({ command: "aidimension add button card" })

Resources#

Beyond tools, the server also exposes resources the agent can read:

Prompts#

The server includes pre-built prompts the agent can call:

• build-a-page — Interactive wizard for new pages
• audit-accessibility — Scan a page for a11y issues
• migrate-from-shadcn — Convert shadcn components to aidimension
• explain-component — Generate docs for a single component

Authentication#

The public MCP server is unauthenticated. For private/team registries, set:

export AIDIMENSION_REGISTRY=https://registry.yourcompany.com
export AIDIMENSION_TOKEN=azr_...

The server will use your token on every request.

Performance#

• First call: ~200ms (cold start, registry fetch)
• Cached calls: ~5ms
• Component copy: ~50ms (write + format)
• Magic build: 3-10s (depends on LLM)

Versioning#

The MCP server reports its version in the initialize handshake:

{
  "protocolVersion": "2025-06-18",
  "serverInfo": {
    "name": "aidimension-mcp",
    "version": "0.4.1"
  }
}

Pin a specific version with @aidimension/mcp@0.4.1 to avoid breaking changes.

Source#

The MCP server is open source at github.com/pathtoresiliencebv/aidimension-mcp. Issues and PRs welcome.