⚡ Wander AI (wanderai)

A dual-mode interactive CLI and compliant Model Context Protocol (MCP) server for multi-agent software development teams.

🚀 Features

• Dual-Mode Architecture:
  • CLI Mode: Command-line tool to initialize a Wander AI multi-agent workspace with agent configurations (.agents/example_ui.agent.md) and default environment configurations (.env.example).
  • MCP Mode: A fully compliant Model Context Protocol (MCP) server communicating over stdio transport, enabling immediate connection with Claude Desktop, Cursor, or any other MCP host.
• Orchestration Bridge: Contains a tool synthesize_multi_agent designed to delegate prompts and instructions to an underlying multi-agent setup, routing requests between Solutions Architect, Tech Lead, Full Stack Developer, and more.
• DX Polish: Built-in mcp-config helper to quickly print clean JSON config blocks for seamless copy-pasting.

📦 Installation

To use wanderai globally in your terminal or as an MCP server, install it globally from the NPM registry:

npm install -g wanderai

🛠️ Usage

Mode 1: Interactive CLI Tool (Default)

1. Initialize a Workspace

To scaffold a new workspace in your current directory (creates .agents/ folder and .env.example):

wanderai init

2. Get MCP Configuration Snippet

To print the JSON snippet needed to connect this tool globally to your MCP host (like Claude Desktop or Cursor):

wanderai mcp-config

Mode 2: Model Context Protocol (MCP) Server

Launch the Model Context Protocol server over stdio:

wanderai mcp

Note: During MCP mode, stdout is reserved strictly for protocol JSON-RPC payloads. All debugging, trace logs, and execution highlights are safely routed to stderr using our Chalk-enhanced logging pipeline to guarantee a zero-pollution transport layer.

Registered Tools

• synthesize_multi_agent: Triggers the Wander AI multi-agent orchestration pipeline to synthesize code or execute an agentic task.
  • Arguments:
    • task_description (string, required): A clear description of the development task you want the multi-agent system to execute.

⚙️ Host Integration Examples

1. Claude Desktop

Add the following snippet inside your Claude Desktop configuration file (typically located at %APPDATA%\Claude\claude_desktop_config.json on Windows or ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "wanderai": {
      "command": "wanderai",
      "args": ["mcp"]
    }
  }
}

2. Cursor

Go to Settings > Features > MCP, click + Add New MCP Server, and configure it with:

• Name: wanderai
• Type: stdio
• Command: wanderai mcp

📂 Directory Structure

Once initialized, your project workspace will look like this:

your-project-workspace/
├── .agents/
│   └── example_ui.agent.md   # Base prompt and role definition for your Frontend Agent
└── .env.example              # Template keys for OpenAI, Gemini, and OpenRouter API integrations

📄 License

MIT © 2026 Wander AI Team.