Hanzo

You run a unified MCP server that exposes a rich set of AI development tools, UI components, and orchestration capabilities. It’s designed to be modular, letting you enable only the toolsets you need while coordinating multiple AI agents and workflows from a single, cohesive platform.

How to use

To use the MCP server, you connect a client that understands MCP tooling and can drive the server to perform actions like file operations, code editing, UI component management, and distributed agent orchestration. You can start with core tools for baseline functionality or enable broader tool categories to match your development workflow. Your client will issue tool-enabled requests, and the server will respond with results or progress.

How to install

Prerequisites you need before installing:
- Node.js (for the CLI and tooling)
- npm (package manager that comes with Node.js)
- Optional: a MCP client you will use to connect to the server (CLI or SDKs)

Additional setup and configuration

# Global install to provide the MCP server CLI
npm install -g @hanzo/mcp

# Start the MCP server with all tools enabled by default
hanzo-mcp serve

# Start the MCP server with core tools only
hanzo-mcp serve --core-only

# Alternatively, configure client-side usage examples (see below) to connect with your MCP client

Notes for using with MCP clients

You can connect using different client configurations. For local CLI usage, you run the server directly and issue commands. For editor or desktop integrations, you configure a client profile to run the server via a package runner like npx, which forwards the serve command to the MCP runtime.

Troubleshooting and tips

If you encounter issues starting the server, ensure Node.js and npm are correctly installed, and that there are no port conflicts. Check that the MCP CLI is globally accessible in your shell and that you have network access if you are using remote clients. If a particular tool category is not responding, verify that category is enabled in your start command (for example, --core-only to limit to core tools or --disable-ui to turn off UI tools).

Available tools

file_operations

Core file operations including read, write, list, create, delete, and move files.

search

Unified search facilities such as grep and file searches across projects.

editing

Single and multi-edit operations for source files.

shell

Execute shell commands and manage background processes.

component_management

List, search, and retrieve details for UI components.

source_code_access

Access component source and demos for quick reference.

project_setup

Initialize projects leveraging Hanzo UI patterns and templates.

blocks_patterns

Access a library of UI blocks and design patterns.

installation_guides

Provide complete setup documentation and guides for projects.

mouse_control

Automate mouse movements, clicks, drags, and scrolling.

keyboard_control

Automate typing, keystrokes, and hotkey combinations.

screen_capture

Capture screenshots and perform pixel-level checks.

image_recognition

Identify on-screen images with confidence metrics.

window_management

List, activate, and control application windows.

multi_monitor

Gather screen information and coordinates across multiple displays.

agent_spawning

Create AI agents with specific models and constraints.

swarm_orchestration

Coordinate multiple agents in parallel or sequential steps.

critic_agents

Code review and automated quality gates for outputs.

hanzo_node

Connect to distributed Hanzo compute nodes.

llm_router

Intelligent model selection with fallback chains.

consensus

Multi-agent voting and decision-making processes.