mcp

The @hanzo/mcp server provides a unified interface for AI development tooling and UI component management through the MCP protocol. By starting the server you enable a collection of core tools (file operations, search, editing, and shell execution) along with optional UI tools, AutoGUI capabilities, and orchestration features that can be toggled on or off. This modular toolset is designed to support AI workflows, automation tasks, UI component retrieval, project setup, and distributed computation coordination. You can access the full suite via the CLI commands or programmatic API, and connect AI clients (like Claude Desktop or Cursor) through the MCP configuration.

You can list available tools, run the full toolset, or selectively enable categories (core, UI, AutoGUI, orchestration) depending on your needs. The server exposes both a CLI experience for quick local usage and a programmatic API for integrating custom workflows. Tools such as file operations, search, code editing, GUI automation, screen/ image handling, and multi-agent orchestration are available, enabling complex AI-assisted development and robotic-like automation within a single MCP server instance.

Prerequisites:

• Node.js (recommended LTS) and npm installed on your system
• Basic familiarity with CLI usage

Installation steps:

1. Install the MCP server globally via npm npm install -g @hanzo/mcp

2. Verify installation hanzo-mcp --version

3. Start the server with default toolset (core + UI + AutoGUI + orchestration by default) hanzo-mcp serve

4. (Optional) Run with only core tools hanzo-mcp serve --core-only

5. (Optional) Disable specific categories hanzo-mcp serve --disable-ui --disable-autogui --disable-orchestration

6. List available tools hanzo-mcp list-tools

7. Install for Claude Desktop (example) hanzo-mcp install-desktop

Tips:

• The server exposes 54 tools by default; you can customize via command flags to enable/disable categories or specific tools.
• Tool configuration supports programmatic usage via createMCPServer for integration into your own applications.
• If you plan to connect external MCP clients (e.g., Claude Desktop, Cursor), add appropriate entries in your .mcp.json or workspace configuration referencing the command and arguments to start the server (as shown in the README examples).
• Ensure network/firewall settings allow MCP client connections if you run the server on a remote host.
• Common issues often relate to version mismatches between the MCP client and server; align client versions with the installed server package.