AIW - AI Workflow Orchestration Tool
Universal AI CLI Management Platform with Intelligent MCP Routing & Transparent Parameter Forwarding
AIW is an intelligent platform for managing AI CLI tools (Claude, Codex, Gemini) with MCP routing, process tracking, provider management, and transparent parameter forwarding.
๐ฏ Core Features
1. AI CLI Management
- Process Tree Tracking: Monitor AI CLI processes and their children
- Provider Management: Switch between AI providers (OpenAI, Anthropic, Google, etc.)
- Transparent Parameter Forwarding: Seamlessly pass all CLI parameters while using provider management
- Capability Detection: Auto-detect installed AI CLI tools
- Wait Mode: Block until AI CLI tasks complete
# Basic usage with provider management
# ๐ Transparent Parameter Forwarding (v0.5.23+)
# Interactive mode with parameter forwarding
# Wait for all AI CLI tasks to complete
# Manage providers (TUI interface)
5. Update (ๆดๆฐAIWๅAI CLIๅทฅๅ ท)
- AIW Self-Update: Update AIW itself to latest version from NPM
- AI CLI Tools Update: Update all installed AI CLI tools (claude, codex, gemini)
- Combined Update: Update both AIW and AI CLI tools in one command
# Update both AIW and all AI CLI tools
# Command will:
# 1. Check and update AIW to latest version
# 2. Check and update all installed AI CLI tools
# 3. Show detailed results for both updates
2. Intelligent MCP Routing โญ
Route user requests to the best MCP tool with 98% token reduction.
How it works:
- LLM ReAct Mode: Uses LLM reasoning to analyze requests
- Vector Search Mode: Semantic similarity fallback
- Dynamic Registration: Registers tools on-demand (98% token savings)
- JavaScript Orchestration: Auto-generates workflows for complex tasks
Usage:
# Start MCP server
# Configure in Claude Code (~/.claude/settings.json)
{
}
# Available MCP tools:
# - intelligent_route: Auto-route to best tool
# - get_method_schema: Get tool schemas
# - execute_tool: Execute tools with negotiation
3. MCP Server Management
Manage external MCP servers with hot-reload support.
# Add MCP server
# List servers
# Enable/disable servers
# Hot-reload: Changes apply instantly without restart
4. MCP Registry CLI โญ
Search and install MCP servers from multiple registries with multi-source aggregation.
Supported Registries:
- Official MCP Registry (registry.modelcontextprotocol.io)
- Smithery (registry.smithery.ai)
# Interactive browse all MCP servers (with fuzzy search)
# Search MCP servers across all registries
# Get detailed server information
# Install MCP server with interactive setup
# Install with environment variables
# Update registry cache
Features:
- Interactive Browse: Fuzzy search through all servers with โโ navigation
- Multi-source Aggregation: Search across Official Registry + Smithery in parallel
- Deduplication: Same server from multiple sources shown once with best match
- Interactive Install: Configure environment variables during installation
- Source Tracking: Track where each server was installed from
๐ Transparent Parameter Forwarding (v0.5.23+)
AIW now supports transparent parameter forwarding, allowing you to use all AI CLI features while maintaining provider management capabilities.
How It Works
# Parameter parsing logic:
# ^^^^^^^^ ^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^
# AIWๅค็ AIWๅค็ ้ๆ่ฝฌๅๅฐAI CLI
Rules
- AIW Consumes:
-p/--providerparameters for provider selection - Transparent Forwarding: All other
-prefixed parameters are passed directly to the AI CLI - Parameter Order: Provider flags must come before other CLI parameters
- Full Compatibility: Maintain complete access to all AI CLI features
Examples
Task Mode
# Model selection with debugging
# Structured output
# Tool restrictions
# Multiple parameters
Interactive Mode
# Interactive session with custom model and debugging
# Interactive with output formatting
# Interactive with specific tools
# Multi-AI with same provider and parameters
Provider-Specific Examples
# Claude with structured output
# Codex with custom settings
# Gemini with approval mode
Benefits
โ Full CLI Access: Use all AI CLI parameters without limitations โ Provider Flexibility: Switch providers without changing commands โ Process Tracking: Maintain AIW's monitoring and task management โ Environment Injection: Automatic provider configuration injection โ Zero Learning Curve: Works exactly like native AI CLI with provider prefix
๐ Quick Start
Installation
# Install from NPM
# Or use npx directly
Initial Setup
# Verify installation
# Check available AI CLI tools
# Configure MCP servers
๐ Documentation
- MCP Usage Guide - Detailed MCP configuration and usage
- GitHub Repository - Full source code and issues
๐ ๏ธ Configuration
Environment Variables
# AI CLI Configuration
# claude, codex, or gemini
# Any provider from provider.json
# MCP Server Configuration
Configuration Files
- MCP Servers:
~/.aiw/mcp.json(standardmcpServersschema compatible with Claude Code)- Claude Code users: Add to
~/.claude/settings.jsonunder the "mcpServers" section
- Claude Code users: Add to
- Providers:
~/.aiw/providers.json
๐ Acknowledgments
- Claude Code: Anthropic's official CLI for Claude
- Anthropic Code Execution: Inspired by Code Execution with MCP
๐ License
MIT License - see LICENSE file for details.
AIW - Universal AI CLI Management Platform with MCP Routing v0.5.22
For full documentation and source code, visit: https://github.com/putao520/agentic-warden