Umbraco MCP

An MCP (Model Context Protocol) server for Umbraco CMS that unlocks AI-powered content management. It provides comprehensive access to the Umbraco Management API, enabling your AI agent to perform back office tasks through natural conversation - freeing you from the UI and making complex, repetitive workflows that would be tedious or impossible manually become effortless.

Intro

The MCP server authenticates using an Umbraco API user, ensuring secure, permission-based access to the Umbraco Management API. This means you maintain complete control over what your AI agent can do through Umbraco's standard user permission system - no special security model required.

Quick Start

1. Create an Umbraco API User

First, create an Umbraco API user with appropriate permissions. You can find instructions in Umbraco's documentation.

2. Install in Claude Desktop

Download and install the Claude Desktop app, then add the MCP server to your configuration:

1. Open Claude Desktop Settings > Developer > Edit Config
2. Add this configuration to your claude_desktop_config.json:

{
  "mcpServers": {
    "umbraco-mcp": {
      "command": "npx",
      "args": ["@umbraco-cms/mcp-dev@17"],
      "env": {
        "NODE_TLS_REJECT_UNAUTHORIZED": "0",
        "UMBRACO_CLIENT_ID": "your-api-user-id",
        "UMBRACO_CLIENT_SECRET": "your-api-secret",
        "UMBRACO_BASE_URL": "https://localhost:{port}",
        "UMBRACO_INCLUDE_TOOL_COLLECTIONS": "document,media,document-type,data-type"
      }
    }
  }
}

3. Fully restart Claude Desktop (including in the system tray on Windows)

Tip: Pin @umbraco-cms/mcp-dev to your site's Umbraco major (e.g. @17 above, or @18 for an Umbraco 18.x site) rather than @latest. Tool schemas track the Management API of the package's own major, so a mismatched major can fail confusingly mid-task once a tool's shape has drifted. The server checks the connected site's version on startup and warns on a mismatch, but pinning avoids it altogether.

Documentation

For complete installation instructions, configuration options, tool listings, and usage examples, see the full documentation:

Umbraco MCP Documentation

Hosted Deployment

The Cloudflare Worker entry point (src/worker.ts) supports two deployment modes, selected at request time by the @umbraco-cms/mcp-hosted library:

• Single-tenant (default) — the worker authenticates against the UMBRACO_BASE_URL configured in wrangler.toml. This is the existing behavior and requires no extra configuration.
• Multi-tenant Umbraco Cloud — set UMBRACO_CLOUD_ROUTING_ENABLED = "true" under [vars] in wrangler.toml. Requests to /at/{alias}/... resolve the upstream Umbraco Cloud project for {alias} and issue access tokens scoped to that resource per the MCP resource-indicator spec. Each Umbraco Cloud project served by the worker must register an OpenIddict client with id umbraco-cms-developer-mcp-hosted.

siteRouting is wired unconditionally; the per-request toggle is the env var.

Contributing with AI Tools

This project is optimized for development with AI coding assistants. We provide instruction files for popular AI tools to help maintain consistency with our established patterns and testing standards.

Using rulesync

The project includes rulesync configuration files that can automatically generate instruction files for 19+ AI development tools. Generate configuration files for your preferred AI tools:

# Generate only for Claude Code
npx rulesync generate --claudecode

# Generate only for Cursor
npx rulesync generate --cursor

# Generate only for Vs Code Copilot
npx rulesync generate --copilot

Other AI Tools

rulesync supports 19+ AI development tools including GitHub Copilot, Cline, Aider, and more. Check the rulesync repository for the complete list of supported tools.

The instruction files cover:

• MCP development patterns and conventions
• TypeScript implementation guidelines
• Comprehensive testing standards with builders and helpers
• Project-specific context and architecture
• API integration patterns with Umbraco Management API